Sclerotherapy for the management of oesophageal varices was described initially by Crafoord and Freckner[17] in 1939 with injection of Quinine. However, it was not until later in the 20th century that this therapy became commonplace in the management of variceal haemorrhage, especially with the advent of fibre-optic endoscopy. Surgery was the mainstay of therapy for variceal haemorrhage prior to the 1970s. Surgical techniques such as oesophageal stapling or oesophagectomy were used, but with high mortality rates from complications such as sepsis, liver failure and renal failure[18]. In patients with portal hypertension, devascularisation procedures were shown to reduce variceal bleeding and mortality in primary prophylaxis in the 1980s, although there was heterogeneity in one such study by Inokuchi et al[19] with recruitment from a total of 22 centres. Shunt formation such as a splenorenal shunt was also performed with rebleeding rates varying from 5%-40%[20,21]. The role of splenectomy was and continues to be useful in patients with segmental portal hypertension secondary to an isolated splenic vein thrombosis. However, this surgical procedure was established later in the 20^th century. Surgical therapies in present guidelines are reserved for patients who fail endoscopic therapies, and have been superseded by either TIPSS as rescue therapy or early TIPSS post index variceal bleeding, which will be discussed later[2].

Another method used prior to the advent of endoscopic therapy pre-1970s was balloon tamponade. The Sengstaken-Blakemore tube’s use was first described in 1950 by Sengstaken and Blakemore[22] although the role of balloon tamponade was initially described in 1930[23]. Its place has largely been superseded by endoscopic therapies, however 21^st century guidelines[24] still suggest a role for balloon tamponade, being used in massive haemorrhage as a bridge until definitive treatment can be instituted (for a maximum of 24 h). Although developed pre-1970s, its role in variceal haemorrhage was secured later in the century with effectiveness in controlling acute bleeding in up to 90% of patients, however with up to 50% rebleeding rates when the balloon was deflated[25]. Complications of balloon tamponade include aspiration pneumonia (often compounded by variceal haemorrhage event itself in encephalopathic patients) and oesophageal ulceration or rupture[26] in up to 15%-20%.

The Linton-Nachlas balloon was developed in the 1970s[27] with a single 600 mL gastric balloon. The safety of this tube compared to Sengstaken Blakemore tube was identified in controlled trial of 79 patients with oesophagogastric variceal haemorrhage[28]. Both types of tamponade therapies resulted in primary haemostasis rates of 86%, but when bleeding from oesophageal varices was assessed, the Sengstaken Blakemore tube achieved permanent haemostasis in 52% compared the Linton-Nachlas tube 30%. The latter was more effective at controlling gastric variceal haemorrhage with 50% primary haemostasis rates compared to total failure in the Sengstaken Blakemore arm. The use of balloon tamponade as definitive therapy however was to be revolutionised by the advent of the fibre-optic endoscope and the therapies that could be delivered with it.

Rigid endoscopes were replaced by narrow fibre-optic endoscopes allowing therapy to be deployed through accessory channels. With a new and easier method for not only diagnosis of variceal haemorrhage but also therapeutic manoeuvres, new therapies were developed. The use of the overtube was phased out, patient comfort was improved and twin channel endoscopes were developed. The first reported case series of endoscopic sclerotherapy[29] was published in the early 1970s with its use becoming more established in the 1980s and thereafter. The concept was that the bleeding varix would “thrombose off” by internal injection of sclerosant causing vascular thrombosis and vascular obliteration[30]. Ethanolamine oleate, sodium tetradecyl sulphate, polidocanol, sodium morrhuate and ethanol have been used for injection sclerotherapy and successfully used in controlled trials[31]. In Europe the most common agents used were ethanolamine oleate and polidocanol, whereas in the United States sodium morrhuate was preferred[32,33]. Paravariceal injection involved injection around the varix causing variceal occlusion by tamponade and subsequent submucosal fibrosis of tissue around the varix, whereas intra-variceal injection induced thrombosis and subsequent occlusion of the lumen[34]. A meta-analysis by D’Amico et al[1] showed the type and volume of sclerosant did not seem to affect the efficacy.

Another issue of trials using injection sclerotherapy in the late 1980s (and 1990s) was the confounding factor that some trials had patients who were not actively bleeding at the time of initial endoscopy[1,35]. Furthermore, the optimal doses of sclerosing agents is unknown, with heterogeneity in scheduled follow-up endoscopies, and also differences between para- and intra-variceal injections[36,37]. There was however no doubt of sclerotherapy efficacy in the role of variceal bleeding. Sclerotherapy was compared to placebo in a controlled trial, with 56 patients having sclerotherapy injection and 60 placebo in patients with variceal bleeding. Survival was significantly better in those treated by sclerotherapy (P < 0.001)[38]. Sclerotherapy was also compared with oesophageal transection in 4 randomised trials[39-42] with similar mortality rates but rebleeding rates higher in the sclerotherapy arms. Only one trial showed a statistically significant reduction in failure to control bleeding with surgery[39]. When sclerotherapy was compared to balloon tamponade in 4 trials[43-46], 2 trials showed significantly higher control of bleeding with sclerotherapy[43,44].

In 1988, the first human cases were described of the use of EVBL in patients with oesophageal varices, based on the concept of banding haemorrhoids with elastic O-rings[47]. This technique was initially applied to canine models in the late 1980s[48,49] and then to patients with portal hypertension by Van Stiegmann et al[50]. EVBL was then successfully incorporated into the management of oesophageal variceal bleeding in the 1990s.

In the 1990s, further trials were carried out with injection sclerotherapy in not only oesophageal but also gastric variceal haemorrhage. Endoscopic therapy with sclerotherapy was found to control active bleeding from oesophageal varices in more than 90% of patients, and effective in reducing the frequency of rebleeding[51-53]. Injection sclerotherapy agents were compared, however most studies found them to have similar efficacy, although with some differences in cost[54,55] and time to obliteration[54,56]. The choice of sclerosant was dependant often on the operator and availability in the endoscopy units. A meta-analysis of 5 studies (Laine L, personal communication[24]) of 251 patients, showed significant benefits of sclerotherapy in terms of initial haemostasis rates compared to sham sclerotherapy, vasopressor therapy alone or balloon tamponade. In another meta-analysis, sclerotherapy was found to be the “gold standard” in acute variceal bleeding[57] with survival benefit seen when used in combination with vasoconstrictors than vasoconstrictors alone. Thus its role in the management of variceal bleeding became established. Injection sclerotherapy use was also extended to the treatment of gastric varices initially by Gotlib and Zimmerman[58] in 1984. The mechanism of action became clearer in the 1980s and 1990s with reports of gastric variceal endothelial damage with subsequent sclerosis[58]. Sarin et al[59] reported a 71.6% variceal obliteration rate in patients with gastric variceal haemorrhage treated with sclerotherapy. However, high re-bleeding rates of up to 60%-90% were reported[26,60]. The combination of ethanolamine sclerosant and cyanoacrylate glue was reported to produce rapid eradication of oesophagogastric varices, with fewer number of injection sessions[61,62].

The 1990s also saw the role of EVBL developed to the forefront of oesophageal variceal haemorrhage. EVBL however is not without complications including: oesophageal ulceration, chest pain, transient dysphagia and occasionally oesophageal stricturing seen at follow-up endoscopy. EVBL however evolved in the 1990s and into the 21^st century as the recommended standard treatment for bleeding oesophageal varices[24]. In a meta-analysis of 10 randomised controlled trials comparing sclerotherapy with EVBL, there was a non-significant benefit of EVBL in achieving initial haemostasis vs sclerotherapy (pooled relative risk of 0.53 with 95%CI: 0.28-1.01)[63]. In one particular study, HVPG increased significantly immediately after both therapies but remained elevated for the duration of the 5 d in the sclerotherapy group whilst returning to baseline levels by 48 h after EVBL group[64] thus potentially identifying a rationale for the use of EVBL over sclerotherapy. In another meta-analysis there was no difference in initial haemostasis rates between both modalities (RR = 1.1, 95%CI: 0.4-2.9)[65], but actively bleeding patients represented only a small subset from larger trials[24].

To complement endoscopic therapies, pharmacological therapies were developed for optimising outcomes in variceal bleeding (Table 3). The lowering of portal pressure, even prior to endoscopy, if the source of upper gastrointestinal bleeding was suspected to be variceal[2,66] became an important issue. To that end, vasopressin and terlipressin were developed and deployed in such a setting. Terlipressin (triglycyl-lysine vasopressin) is a synthetic analogue of vasopressin with longer half-life negating the need for continuous infusion and acts on V1 receptors leading to splanchnic vasoconstriction. This in turn reduces portal inflow and pressure. Consequently there is an improvement in renal blood flow and reduction in portal pressure. Blockade of the V2 receptors can also result in free water absorption in the renal collecting ducts. Vasopressin (mainly used in the United States due to the unavailability in terlipressin) had been shown to achieve haemostasis in 60%-80%[67] of patients, but compared with terlipressin had less effect on the reduction of early rebleeding and did not improve survival from active variceal haemorrhage. Terlipressin was shown to reduce all-cause mortality when compared to placebo in meta-analyses[68,69] and guidelines recommend early treatment, which should be continued for up to 5 d[24] when potential for rebleeding is greatest. Side-effects include peripheral or coronary ischaemia, nausea and diahorrea. Blocking activation of the V2 receptors of the renal tubules can cause a dilutional hyponatraemia, an effect that reverses rapidly on discontinuation of the drug. When compared to somatostatin analogues such as octreotide, the haemodynamic effects of terlipressin on portal pressure were found to be more sustained[69] suggesting terlipressin might have a more prolonged benefit in bleeding varices. Thus vasoactive drugs became a key part of the initial therapy in variceal haemorrhage.

One of the major radiological advances in the management of variceal haemorrhage in the 1990’s was the advent of TIPSS. Although first described in 1983 by Colapinto et al[70], it was largely in the 1990s and thereafter that its place in the management of portal hypertensive complications was secured. TIPSS involves the placement of a stent between the portal and hepatic vein to reduce portal pressure, thus stemming variceal haemorrhage or preventing rebleeding. Complications of TIPSS include haemorrhage, infection, intravascular haemolysis, liver dysfunction, shunt dysfunction and worsening of hepatic encephalopathy[71,72]. Initial TIPSS were bare-metal stents with rebleeding rates of up to 20% at 2 years[73]. TIPSS was initially used for uncontrolled bleeding with control of bleeding in 90%-95% of patients and a 4-wk survival of 50%-60%[74]. In a review of 15 studies, immediate haemostasis rates of 93% were found with rebleeding rates of 12%[75]. In another meta-analysis of 11 randomised controlled trials, although TIPSS reduced risk of rebleeding, TIPSS was found to not affect survival in patients with variceal haemorrhage[76]. TIPSS was also found to be successful in the management of bleeding gastric varices[77-79].

With the dawn on the 21^st century, pharmacological, endoscopic and radiological therapies for variceal haemorrhage improved outcomes. The role of antibiotics in variceal bleeding became clear in the early 21^st century. Primary or secondary bacterial infections are common in cirrhotic patients[80,81] due to bacterial translocation into the portal system from impaired mucosal integrity and an impaired immune function. Antibiotics were found to reduce bacterial infections, recurrent bleeding and improve mortality in patients bleeding from oesophageal varices[82-84]. Current guidelines recommend broad-spectrum antibiotic prophylaxis[2,24,26] in patients with suspected and proven variceal haemorrhage. Local antibiotic policy can vary and a patient’s “nil-by mouth” status can influence the choice of antibiotic. However, oral quinolones are recommended, or a 3^rd generation cephalosporin in patients who received quinolone prophylaxis, have advanced cirrhosis, or live in areas of high quinolone resistance[2]. Another area of interest recently in resuscitation has been that of transfusion. In a study by Villanueva et al[85] the role of over-transfusion in GI bleeding has been explored and its effects on portal pressure. In patients with a liberal transfusion strategy [transfused when haemoglobin (Hb) fell to less than 9 g/dL] there was a significant rise in portal pressure gradient in the 1^st five d post bleed compared to patients with a restrictive transfusion strategy (transfused when Hb fell to less than 7 g/dL). Thus is could be argued that patients with a variceal bleed are not as aggressively resuscitated/over-transfused as they may have previously been, however further clarification in this area is required.

Endoscopic therapy developed further in the 21^st century, with obturation therapies for gastric variceal bleeding coming of age in the new millennium. Gastric varices account for 10%-30% of variceal haemorrhage, and although less common than oesophageal varices, when bleeding occurs it can often be torrential and associated with a high mortality[86-89]. Gastric varices can also bleed at a lower portal pressure than oesophageal varices[86-89]. There is limited data on EVBL in the management of gastric variceal bleeding with high rates of gastric variceal recurrence following EVBL due to a more superficial effect compared with obturation therapy[86]. Technical difficulty of banding in a retroflexed endoscope position and a theoretical risk of gastric rupture has meant EVBL for gastric varices has largely been superseded by obturation therapies using thrombin and N-butyl-2-cyanoacrylate (Histoacryl ™) injection. N-butyl-2-cyanoacrylate is a long-chain cyanoacrylate glue that polymerises and solidifies within seconds following contact with blood in a gastric varix. It is mixed with the oily agent Lipiodol delaying polymerisation. Complications of its use include: endoscope damage due to blockage of the injection channel, sticking of the injection needle into a varix, mediastinitis, local abscesses, and cerebral/pulmonary embolisms or splenic infarcts from glue migration. Immediate haemostasis rates of 92%-100% have been reported with variable re-bleeding rates[86-93]. Cyanoacrylate glue when compared with ethanol injection in a randomised study had faster rates of variceal obliteration, required smaller injection volumes, had improved efficacy in control of acute gastric variceal bleeding and reduced need for rescue surgery[92,93]. Thrombin was another obturation therapy developed in the 21^st century used in acutely bleeding gastric varices. It is a haemostatic agent converting fibrinogen to a fibrin clot and causing platelet aggregation[94]. Initially in the late 1980s and 1990s there were small case-series of its use with haemostasis rates between 70%-100% using bovine thrombin[95-99]. Bovine thrombin was discontinued because of the potential risk of prion transmission. Thus, short-term small uncontrolled studies of human-derived thrombin have demonstrated initial haemostasis rates of 100% but often a high mortality from re-bleeding[99-101]. A recent retrospective study from Edinburgh, United Kingdom demonstrated in 33 patients treated with human thrombin for gastric variceal bleeding rebleeding rates of 10.8%[102]. It is worth noting to date there have been no controlled trials with thrombin vs other obturation treatments such as N-butyl-2-cyanoacrylate) to our knowledge.

Radiological interventions in variceal haemorrhage improved in the new century too. In 2004, the advent of covered TIPSS stent (with an expanded polytetrafluoroethylene cover) was hailed as a breakthrough and approved by the United States Food and Drug Administration. The covered stent improved shunt patency by reducing tissue ingrowth by minimising transmural bile permeation[103]. The primary patency of covered stents at 1-year were found to be up to 80%-90%[104-107] with reduction of rebleeding post “index bleed” to less than 10%[105,107,108]. Other studies confirmed the role of a rescue TIPSS in variceal bleeding which could not be controlled by endoscopy or vasoactive drugs[67,68]. The early TIPSS placement has been shown to have beneficial effect in patients with a HVPG > 20 mmHg presenting with a variceal bleed[11]. In this study published in 2004, patients who were considered high risk (HVPG > 20 mmHg) were selected and randomised to early uncovered TIPSS or standard of care within 24 h of presentation. Treatment failure was deemed as failure to control acute variceal bleeding and/or early rebleeding after the first endoscopic therapy. TIPPS reduced rebleeding and treatment failure, and was associated with superior in-hospital and 1-year survival. However, the therapy used in the control arm was endoscopic sclerotherapy alone, which is not the accepted standard of care. The other major issue translating this study into real world practice was the availability of HVPG in routine clinical practice.

Interventional radiological procedures for the treatment of gastric varices in the 2000s included the advent of BRTO[109-111] as salvage or rescue therapy when endoscopic obturation therapy fails. BRTO is an interventional radiological technique for gastric variceal bleeding whereby a splenorenal shunt often seen in such patients can be occluded with sclerosant using a balloon catheter approached via the left renal vein[109-111]. BRTO may potentially become an alternative to TIPSS in patients with active gastric variceal bleeding in whom a gastrorenal shunt is present[110]. However, it is not commonly used out-with the Far East or large tertiary referral centres. There is also an increased risk of the development of oesophageal varices after its use[109]. Its role has not been incorporated in any European or United States guidelines to date[2,24].

Liver transplantation is the only curative treatment for liver cirrhosis at this point in time, although its role in bleeding variceal haemorrhage has not been established. In some centres it has been proposed as a treatment in patients with advanced liver disease who fail endoscopic therapies[112]. These studies were however in the era prior to EVBL, combined pharmacological and endoscopic therapies and TIPSS. In a trial by Orloff et al[113] unselected consecutive patients with advanced cirrhosis and bleeding oesophageal varices were studied who had either sclerotherapy (n = 106) or emergency direct portocaval shunt (n = 105). The 3-, 5-, 10- and 15-year survival rates were significantly higher in the portocaval shunt group (P < 0.001). On the follow-up, 6% of patients were referred for liver transplant assessment, 3% listed and only a total of 2% actually underwent liver transplantation for progressive liver failure. A conclusion drawn from the study authors was that transplantation was infrequently required in this setting (even prior to the TIPSS era) and if initial bleeding was controlled (in 100% of the portocaval shunt arm) then survival was similar or better than that following transplant. It should be remembered that such centre-specific data however often differs from centres with less experience in portocaval shunts. However to our knowledge there are no randomised trials of endoscopic therapy with radiological therapy and liver transplantation in the setting for acute variceal haemorrhage and this is certainly not current practice. It could be argued that transplantation should only be reserved in those patients whom combined pharmacological and endoscopic therapy fails along with a trial of radiological intervention such as TIPSS or BRTO, or even surgery. However, patients with poor liver function or in whom liver function does not recover should always be considered at an early stage for liver transplant assessment where appropriate based on local scoring systems such as MELD in the United States and UKELD in the United Kingdom.

Areas of recent interest that required future clarification and further study include the early role of TIPSS in variceal haemorrhage, oesophageal stents and new agents for haemostasis.

The exact and optimal role of TIPSS in variceal haemorrhage has been particularly under the spotlight recently. In 2004, the early placement of TIPSS was shown to have beneficial effect in patients with a HVPG > 20 mmHg presenting with a variceal bleed[11]. In a recent seminal multi-centre European study in 2010[4], 63 cirrhotic patients with oesophageal variceal bleeding were treated with vasoactive drugs plus endoscopic therapy and then randomised to one of two treatment arms if they had Child’s C disease or active bleeding and Child’s B disease. The first arm was covered TIPSS within 72 h (“early-TIPSS”), and the second arm continuation of vasoactive drugs for 3 to 5 d followed by non-selective beta-blockers and with long-term EVBL (with the insertion of a TIPSS only if required as a rescue therapy). Rebleeding or failure to control bleeding occurred in only one patient in the “early TIPSS” arm, and in 14 patients in the control arm (P < 0.001). Overall mortality was lower in the “early-TIPSS” group (12 vs 4 patients, P = 0.01) with 1-year survival of 61% in the control group vs 86% in the “early-TIPSS” group (P < 0.001). There was no difference in the incidence of hepatic encephalopathy. A post RCT surveillance study from the same group published last year, aimed to confirm the results in a clinical setting[114] (Table 4). Patients admitted with acute variceal bleeding and high risk of treatment failure (Child C < 14 or Child B plus active bleeding) were thereafter treated with early covered TIPSS (n = 45) or combined pharmacology/endoscopic therapy (n = 30). The patients treated with “early-TIPSS” had lower rates of rebleeding or failure to control bleeding than patients receiving combined therapy (3 vs 15, P < 0.001). There was a tendency also towards reduced mortality in the “early-TIPSS” group (P = 0.056). Criticisms of the “early TIPSS” trial however included that recruitment was prolonged (3 years) to recruit 63 patients via 9 centres[4], with a high exclusion rate (296 patients excluded). The second issue is that of the inclusion of patients with ongoing bleeding following index endoscopy. This might arguably be termed a “rescue” TIPSS and although no studies have been done in this area it is intuitive to suggest that survival would be improved if haemostasis has not been achieved. Thirdly, survival at 1 year with “early TIPSS” was remarkably high (86% vs 61% in the medical management group)[4]. Thus the current Baveno V guidelines[2] suggest considering the “early TIPSS” approach, but clearly further studies are necessary where patients requiring an “early TIPSS” as a rescue therapy are excluded.

Another novel area of interest recently is the use of self-expanding oesophageal stents, which again will require further study to clarify their role in influencing outcome from variceal haemorrhage. The stent acts by applying direct tamponade to the distal oesophageal mucosa and any associated bleeding varices. Such stents were used in a pilot study in 20 patients who failed to achieve haemostasis with pharmacological or endoscopic techniques[115], and achieved 100% immediate haemostasis rates in such a rescue setting. There was a stent migration in 25% of patients in this initial study and 10% of patients died within 5 d. Three other studies have further been published[116-118], with a combined total of 57 patients. Successful stent placement ranged from 90%-100% and control of bleeding ranging between 70%-100%. Stent migration rates varied from 0%-18% with a total of 4 patients rebleeding. Such stents may be a promising option in refractory oesophageal haemorrhage as bridge therapy to definite treatment such as TIPSS. However, randomized controlled trials with comparison to other interventions or even as an adjunct to current standard of care are necessary before they can be considered standard of care. Their mechanism of action would make them comparable to balloon tamponade and there is currently a study group in Barcelona exploring this (NCT01242280). Another United Kingdom study entitled “Effective haemostasis using self-expandable covered mesh-metal oeosphageal stents vs standard endoscopic therapy in the emergency treatment of oesophageal variceal hemorrhage: A multicenter, open, prospective, randomized, controlled study-ISRCTN 98310189” is under way and recruiting. Preliminary data was recently presented in the use of stents compared to balloon tamponade in variceal bleeding refractory to endoscopic and medical therapy. Escorsell et al[119] reported on 28 patients (15 Sengstaken and 13 metal stents) with the intention to treat analysis showing more frequent success of therapy in the stent arm (46% vs 3%, P = 0.005). There was a trend towards better control of bleeding (P = 0.1) and less transfusion requirements (P = 0.08) in the stent arm. Survival rates were comparable (P = 0.4). The authors concluded that oesophageal stents were indeed more effective than balloon tamponade for temporary control of variceal haemorrhage in treatment failures. The stents however do not have a role in gastric variceal bleeding in their current form.

Another new area of interest that has been the development of haemostatic powders/sprays. TC-325 (Hemospray, Cook Technology™) is a granular non absorbable mineral powder used in the management of arterial wounds. It achieves hemostasis by activating platelets and increasing the concentration of clotting factors and also by forming a mechanical barrier over the wall of a bleeding vessel[120] thus forming a mechanical plug at site of bleeding[121]. It contains no proteins from animals or humans. The spray device kit contains an application catheter, a propellant CO₂ canister and also a chamber containing 20 g of powder. Its role has been studied in patients bleeding from peptic ulcers[122].

In a pilot study by Ibrahim et al[123], the use of one such powder TC-325 (was studied in 2 tertiary care referral centres with primary haemostasis rates and rebleeding rates measured). Nine patients with confirmed variceal bleeding had treatment within 12 h of admission, with 21 g of haemostatic powder applied via a catheter in the accessory channel of the endoscope from the cardia up to 15 cm above the gastro-oesophageal junction. There was no rebleeding within 24 h and no mortality at 15 d. Although a small pilot study, further larger trials needed to secure its position in variceal haemorrhage. In another case series, its role in the management of portal hypertensive bleeding was studied in 4 patients - 3 with portal hypertensive gastropathies and 1 portal colopathy[124]. All patients had cessation of bleeding with Hemospray and reduced transfusion requirements thereafter however in 1 patient a complication of viscus perforation was encountered and the patient died shortly after endoscopy - however it was unclear if perforation was secondary to instrumentation during the procedure or the spray itself. Its use has also been studied in small case reports in the management of bleeding gastric varices[125]. This remains a promising area requiring further large trials securing its position in the management of variceal haemorrhage.