1

Upper Gastrointestinal Bleeding

Acute gastrointestinal (GI) bleeding is a potentially life-threatening abdominal emergency that remains a common cause of hospitalization. The incidence of upper gastrointestinal bleeding (UGIB) is approximately 100 cases per 100,000 population per year (Fallah, 2000). Bleeding from the upper GI tract is approximately 4 times as common as bleeding from the lower GI tract and is a major cause of morbidity and mortality. Mortality rates from UGIB are 6-10% overall (Fallah, 2000). An aging patient population with an increasing prevalence of associated medical comorbidities has kept the mortality figures largely unchanged for the past 30 years, despite technologic advances in endoscopy and other minimally invasive procedures (Peter, 1999). The use of various endoscopic techniques, medical therapies, and visceral angiography has progressively diminished the role of surgery in the emergent management of UGIB. Nevertheless, operative intervention still represents the most definitive intervention and remains the final therapeutic option for many bleeding lesions of the upper GI tract. Of patients who develop UGIB, 3-15% require a surgical procedure (Peter, 1999). With the evolution of laparoscopy, surgeons may have a new dimension to the management of GI tract bleeding by developing new techniques performed through smaller incisions. Clinical: The history and physical examination provide crucial information for the initial evaluation of a patient presenting with a GI tract hemorrhage. The history findings can be extremely helpful in determining the location of the GI hemorrhage. Alcohol abuse or a history of cirrhosis should elicit consideration of portal gastropathy or esophageal varices as sources for the bleeding. A history of recent nonsteroidal anti-inflammatory drug (NSAID) abuse should elicit concern about bleeding from a gastric ulcer (Stabile, 2000). Differential diagnoses for UGIB o Gastric ulcer o Duodenal ulcer o Esophageal varices o Gastric varices o Mallory-Weiss tear o Esophagitis o Neoplasm o Hemorrhagic gastritis o Dieulafoy lesion o Angiodysplasia o Hemobilia o Pancreatic pseudocyst o Pancreatic pseudoaneurysm o Aortoenteric fistula UGIB is defined as bleeding derived from a source proximal to the ligament of Treitz. Hematemesis and melena are the most common presentations of acute UGIB, and patients may present with both symptoms. Occasionally, a brisk UGIB

2 manifests as hematochezia. A recent meta-analysis documented the incidence of presenting symptoms in patients with UGIB as follows (Peter, 1999): History and physical examination findings in acute UGIB at presentation o Hematemesis - 40-50% o Melena - 70-80% o Hematochezia - 15-20% o Either hematochezia or melena - 90-98% o Syncope - 14.4% o Presyncope - 43.2% o Symptoms 30 days prior to admission - No percentage available o Dyspepsia - 18% o Epigastric pain - 41% o Heartburn - 21% o Diffuse abdominal pain - 10% o Dysphagia - 5% o Weight loss - 12% o Jaundice - 5.2% As indicated, hematemesis is observed in 40-55% of patients, including patients with coffee-ground emesis. Melena is documented in approximately 7080% of patients, and hematochezia is documented in approximately 15-20%. These clinical signs may also be indicators of the potential source of the GI bleeding, as noted in the following table (Peter, 1999). Table 1. Probable Source of GI Bleeding Within the Gut Clinical Indicator Hematemesis Melena Hematochezia Blood-streaked stool Probability of Probability of Upper GI Lower GI Source Source Almost certain Probable Possible Rare Rare Possible Probable Almost certain Possible

Occult blood in Possible stool

Some prognostic indicators that can be detected from the history and physical examination findings are helpful for developing a scoring system to assess the risk of poor outcome with UGIB. These factors include age, heart rate, systolic blood pressure (SBP) upon admission, orthostatic changes in blood pressure or pulse rate, and the use of any anticoagulants. Assessing the patient for hemodynamic instability and clinical signs of poor perfusion is important early in the initial evaluation to properly triage patients with massive hemorrhage to intensive care unit (ICU) settings. Worrisome clinical signs and symptoms of hemodynamic compromise include tachycardia of more than 100 beats per minute

3 (bpm), systolic blood pressure of less than 90 mm Hg, cool extremities, syncope, and other obvious signs of shock such as ongoing brisk hematemesis or maroon or bright-red stools, which requires rapid blood transfusion. Age older than 60 years is an independent marker for a poor outcome (Peter, 1999). The mortality rate increases to 12-25% in those older than 60 years compared to a mortality rate of less than 10% for patients younger than 60 years (Peter, 1999). The American Society for Gastrointestinal Endoscopy (ASGE) grouped patients with UGIB according to age and correlated age category to risk of mortality. The ASGE found a mortality rate of 3.3% for patients aged 21-31 years, a rate of 10.1% for those aged 41-50 years, and a rate of 14.4% for those aged 71-80 years (Peter, 1999). Patients who present in hemorrhagic shock have an increased mortality rate of up to 30%. Hemorrhage may be classified based on the amount of blood loss, as noted in the following table (American College of Surgeons Committee on Trauma, 1997). Table 2. Estimated Fluid and Blood Losses in Shock Class 1 Blood mL Loss, Up to 750 Class 2 750-1500 Class 3 1500-2000 30-40% >120 Class 4 >2000 >40% >140

Blood Loss,% Up to 15% 15-30% blood volume Pulse bpm Blood Pressure Respiratory Rate Rate, <100 Normal >100 Normal

Decreased Decreased

Normal or Decreased Decreased Decreased Increased 20-30 Mildly anxious 30-40 Anxious, confused Crystalloid and blood >35 Confused, lethargic Crystalloid and blood

Urine Output, 14-20 mL/h CNS/Mental Status Fluid Replacement, 3-for-1 rule Slightly anxious

Crystalloid Crystalloid

This classification scheme aids in understanding the clinical manifestations of hemorrhagic shock. In early class 1 shock, the patient may have normal vital signs, even with a 15% loss of total blood volume. As the percentage of blood volume loss increases, pertinent clinical signs, symptoms, and findings become more apparent.

4 Although early cardiovascular changes occur as blood loss continues, urine output, as a sign of end organ renal perfusion, is only mildly affected until class 3 hemorrhage has occurred. Bornman et al correlated the presence of shock (defined as pulse rate >100 bpm or SBP <100 mm Hg) with the incidence of rebleeding rates after initial nonsurgical intervention. They found that rebleeding (a marker for increased mortality and need for surgery) occurred in 2% of patients without shock, in 18% with isolated tachycardia, and in 48% with shock. Schiller et al determined that SBP is a sensitive clinical marker for helping predict mortality. They correlated mortality rates based on the patient's SBP at the time of bleeding and found mortality rates of 8% for patients with SBP more than 100 mm Hg, rates of 17% for SBP of 80-90 mm Hg, and rates of more than 30% for SBP less than 80 mm Hg. Unless the patient has evidence of shock, orthostatic testing should be performed to assess and document a hypovolemic state. A positive tilt test finding is defined as an SBP decrease of 10 mm Hg and a pulse rate increase of 20 bpm with standing compared to the supine position. The ASGE was able to correlate orthostatic changes with the incidence of mortality (Silverstein, 1981). The mortality rate when orthostatic changes are present is 13.6%, compared to 8.7% when they are absent. Knopp et al studied the use of the tilt test in phlebotomized healthy volunteers and found that a positive tilt test result consistently correlated with a blood loss of 1000 mL. This becomes extremely useful when evaluating patients with class 1 hemorrhagic shock. Lab Studies: The following laboratory tests are the most useful when evaluating patients with UGIB. o Hemoglobin value and type and crossmatch blood: Based on the patient's initial hemoglobin level and clinical assessment of shock, a type and screen or type and crossmatch should be ordered. The patient should be crossmatched for 2-6 units based on the rate of active bleeding. The hemoglobin level should be monitored serially in order to follow the trend. An unstable hemoglobin level may signify ongoing hemorrhage requiring further intervention. o BUN-to-creatinine ratio: The BUN-to-creatinine ratio increases with UGIB. A ratio of greater than 36 in a patient without renal insufficiency is suggestive of UGIB. o Coagulation profile: The patient's prothrombin time (PT), activated partial thromboplastin time, and International Normalized Ratio (INR) should be checked to document the presence of a coagulopathy. The coagulopathy may be consumptive and associated with a thrombocytopenia. A platelet count of less than 50 with active acute hemorrhage requires a platelet transfusion and fresh frozen plasma in an attempt to replete lost clotting factors. The coagulopathy could be a marker for advanced liver disease. The PT is used in calculating the Child-Pugh score (Corson, 2001). Prolongation of the PT based on an INR of more than 1.5 may indicate moderate liver impairment.

5 A fibrinogen level of less than 100 mg/dL also indicates advanced liver disease with extremely poor synthetic function. o Liver function tests: Liver function tests are also needed to calculate the Child-Pugh score. Elevated aminotransferase levels are a result of hepatocellular injury. Increased levels of alkaline phosphatase and gamma-glutamyltranspeptidase are indicative of cholestatic liver disease. o Plasma fibrinogen level o Serum electrolyte values Patients generally require blood transfusions because of hypoperfusion and hypovolemia. Patients with significant comorbid conditions (eg, advanced cardiovascular disease) should receive blood transfusions to maintain myocardial oxygen delivery to avoid myocardial ischemia. One of the criteria used to determine the need for surgical intervention is the number of units of transfused blood required to resuscitate the patient. The more units required, the higher the mortality rate (Larson, 1986). Operative intervention is indicated once the blood transfusion number reaches more than 5 units, as noted in the following table (Larson, 1986). Table 3. Effect of Number of Packed Erythrocyte Transfusions on Need for Surgery and Mortality from UGIB

Number of Units Need for Mortality Transfused Surgery, % Rate, % 0 1-3 4-5 >5 4 6 17 57 4 14 28 43

Medical therapy: Initial management Resuscitation of a hemodynamically unstable patient begins with assessing and addressing the ABCs (ie, airway, breathing, circulation) of initial management. Patients presenting with severe blood loss and hemorrhagic shock present with mental status changes and confusion. In such circumstances, patients cannot protect their airway, especially when hematemesis is present. In these cases, patients are at increased risk for aspiration, which is a potentially avoidable complication that can significantly affect morbidity and mortality. This situation must be recognized early, and patients must be electively, not emergently, intubated in a controlled setting using cricoid pressure. Once the airway is secured, the next step in evaluation is assessing the patient's circulation. Intravenous access must be obtained. Bilateral 16-gauge (minimum) upper extremity peripheral intravenous lines are adequate for volume resuscitative efforts. Poiseuille law states that the rate of flow through a tube is proportional to the fourth power of the radius of the cannula and is inversely

6 related to its length (Corson, 2001). Thus, short, large-bore peripheral intravenous lines are adequate for rapid fluid infusion. A rough guideline for the total amount of crystalloid fluid volume needed to correct the hypovolemia is the 3for-1 rule. Replace each milliliter of blood loss with 3 mL of crystalloid fluid. This restores the lost plasma volume. Patients with severe coexisting medical illnesses, such as cardiovascular and pulmonary diseases, may require pulmonary artery catheter insertion to closely monitor hemodynamic cardiac performance profiles during the early resuscitative phase. Once the ABCs have been addressed, assess the patient's response to resuscitation based on evidence of end organ perfusion and oxygen delivery. An interesting study published by Kaplan et al evaluated whether the physical examination findings alone or in combination with biochemical markers (arterial lactate levels) can help accurately diagnose hypoperfusion. Their study revealed that skin temperature upon physical examination in combination with serum bicarbonate levels correlated well with the level of systemic perfusion. The patients identified as being hypoperfused based on clinical parameters were confirmed by following the serum lactate or mixed venous oxygen saturation (SvO2) and cardiac index. Pulmonary artery catheters may be helpful to guide therapy. Foley catheter placement is mandatory to allow a continuous evaluation of the urinary output as a guide to renal perfusion. This labor-intensive management should be performed only in an ICU setting. Once the maneuvers to resuscitate are underway, insert a nasogastric tube (NGT) and perform an aspirate and lavage procedure. This should be the first procedure performed to determine whether the GI bleeding is emanating from above or below the ligament of Treitz. If the stomach contains bile but no blood, UGIB is less likely. If the aspirate reveals clear gastric fluid, a duodenal site of bleeding may still be possible. In a retrospective review of 1190 patients, Luk et al found that positive NGT aspirate findings were 93% predictive of an upper GI source of bleeding. The ASGE performed a study comparing NGT aspirate findings to the endoscopic findings of the bleeding source (Silverstein, 1981). This study revealed that 15.9% of patients with a clear NGT aspirate, 29.9% of patients with coffee-ground aspirate, and 48.2% of patients with red blood aspirate were documented as having an active upper GI source of bleeding at the time of endoscopy. Based on data collected by the ASGE, realize that an NGT aspirate finding can be negative even in the setting of a large duodenal bleeding ulcer. A recent study correlated mortality with the color of the fluid from the NGT aspirate and the color of the stool (Kupfer, 2000). As shown in the following table, the color of the NGT aspirate can be a prognostic indicator. Table 4. Effect of the Color of the Nasogastric Aspirate and of the Stool on UGIB Mortality Rate Nasogastric Aspirate Color Clear Stool Color Brown red or Mortality Rate, % 6

7 Coffee-ground Brown black Red Red blood Black Brown Red or 8.2 19.1 12.3 19.4 28.7

Surgical therapy: Endoscopy The development of endoscopy has provided clinicians with the ability for both diagnostic and therapeutic approaches to bleeding from the GI tract. Endoscopic examination of the upper GI tract provides useful information regarding the source and site of bleeding (Silverstein, 1981). Endoscopic findings and the incidence rate in patients with UGIB o Duodenal ulcer - 24.3% o Gastric erosion - 23.4% o Gastric ulcer - 21.3% o Esophageal varices - 10.3% o Mallory-Weiss tear - 7.2% o Esophagitis - 6.3% o Duodenitis - 5.8% o Neoplasm - 2.9% o Stomal (marginal) ulcer - 1.8% o Esophageal ulcer - 1.7% o Other/miscellaneous - 6.8% Endoscopy should be performed immediately after endotracheal intubation (if indicated), hemodynamic stabilization, and adequate monitoring in an ICU setting have been achieved. Studies performed in the 1970s and early 1980s questioned the role for emergent endoscopy because no significant improvement in patient outcomes was noted (Peter, 1999). These studies were performed during an era before the more advanced therapeutic maneuvers of endoscopy were developed. Since the late 1980s, endoscopic techniques to achieve hemostasis for bleeding ulcers and varices have continued to evolve. Endoscopy is now the method of choice for controlling active ulcer hemorrhage. Over the last 10 years, several randomized clinical trials and meta-analyses have demonstrated and supported that early endoscopic hemostatic therapy significantly reduces rates of recurrent bleeding, need for emergent surgery, and mortality in patients with acute nonvariceal UGIB. Cooper et al studied the effectiveness of performing an early endoscopy within the first 24 hours of an acute UGIB episode (Cooper, 1999). Early endoscopy was associated with reductions in the length of hospital stay, rate of recurrent bleeding, and need for emergent surgical intervention. In a retrospective review involving more than 30,000 cases, Yavorski et al showed that the mortality

8 rates were more than twice as high for those who did not undergo an early endoscopic procedure than those who did undergo an early endoscopic procedure (11.1% vs 5.2%). From 1981-1991, the rate of performing endoscopic procedures in the acute management of UGIB increased from 66.7% to 94%. The endoscopic techniques used to achieve hemostasis have continued to evolve over the last 15 years as newer, more effective therapies have been developed. Various techniques currently available for achieving hemostasis (Stabile, 2000) o Injection of vasoactive agents o Injection of sclerosing agents o Bipolar electrocoagulation o Band ligation o Thermal probe coagulation o Constant probe pressure tamponade o Argon plasma coagulator o Laser photocoagulation o Rubber band ligation o Application of hemostatic materials, including biologic glue The 3 most popular methods of hemostasis are injection therapy, coaptive coagulation, and laser phototherapy. Injection therapy involves the use of several different solutions injected into and around the bleeding lesion. The different solutions available for injection are epinephrine, sclerosants, and clot-producing materials such as fibrin glue. The epinephrine used for injection is diluted (1:10,000) and injected as 0.5- to 1-mL aliquots. Epinephrine is a vasoactive agent that works by inducing vasoconstriction and decreasing blood flow to the area. This allows for increased platelet function and clot formation to attain hemostasis. Debate continues over whether the hemostatic effect of epinephrine is due to the induced vessel vasoconstriction and subsequent platelet aggregation or to the tamponade effect produced by injecting the volume of drug into the tissue surrounding the bleeding lesion. Injecting a volume of sterile isotonic sodium chloride solution and providing a tamponade effect also leads to hemostasis, although not as effectively as epinephrine (Fallah, 2000). Although the injected epinephrine is absorbed into the systemic circulation, this does not appear to have any adverse effects on hemodynamic status. The sclerosant solutions used today include ethanol, polidocanol, and sodium tetradecyl sulfate. The sclerosants create hemostasis by inducing thrombosis, tissue necrosis, and inflammation at the site of injection. When large volumes are injected, the area of tissue necrosis can produce an increased risk of local complications such as perforation. Combining the various agents into a single injection has not been shown to be more beneficial than single-agent therapy alone (Fallah, 2000). Combining epinephrine injections with human thrombin (600-1000 IU) reduces the risk of bleeding (Fallah, 2000). The use of fibrin glue has been shown to be successful, with results similar to that of epinephrine injections (Vargas, 1999).

9 Coaptive coagulation uses direct pressure and thermal therapy to achieve hemostasis. Thermal therapy includes monopolar and bipolar electrocoagulation and heater-probe application. The bleeding vessel is isolated, compressed, and tamponaded prior to coagulation therapy. By using both maneuvers, the depth of tissue injury is minimized. Coaptive coagulation is as effective as injection therapy in achieving hemostasis (Fallah, 2000). Laser phototherapy uses an Nd:YAG laser to create hemostasis by generating heat and direct vessel coagulation. This is a noncontact thermal method. It is not as effective as coaptive coagulation because it lacks the use of compression to create a tamponade effect (Fallah, 2000). An additional deterrent to its use is expense. Although injection therapy and thermal therapy are highly successful methods to control a bleeding vessel, rebleeding still occurs in 15-20% of cases. Combining injection therapy with heater-probe coagulation can be used in an attempt to reduce the rebleeding rate in high-risk patients who have spurting arterial bleeding observed during endoscopy. The epinephrine injection stops the bleeding by vasoconstricting the vessel, which then provides a clearer view for the endoscopist to see the lesion, to apply the heater probe to the appropriate area, and to permanently seal the vessel. Chung et al compared single-agent therapy to combined epinephrine injection with heater-probe application. The study revealed a reduction in the incidence of rebleeding and the need for emergent surgical intervention when using combined therapy. The endoscopic appearance of the bleeding lesion has been used to identify patients at high risk for recurrent bleeding (Chung, 1994). The indication for endoscopic therapy is based on the size, site, and stigmata of recent bleeding. An actively bleeding vessel is treated because it is a high-risk lesion with a 55% risk for recurrent bleeding. A nonbleeding visible vessel is also a high-risk lesion that is treated because of its 43% risk for rebleeding. Recurrent bleeding occurs most commonly within the first 72 hours. Treatment of clots adherent to a lesion is based on the patient's risk factors to withstand and survive another bleeding episode. The risk for rebleeding varies from 14-36%. Low-risk lesions are those that appear as flat, pigmented spots and those that involve a clean ulcer base with no visible vessel. The risk for rebleeding in these lesions is 10% and 5%, respectively. Endoscopic treatment of these lowrisk lesions is not usually performed (Chung, 1994). Peptic ulcer disease Peptic ulcer disease (PUD) remains the most common cause of UGIB. In a literature review involving more than 10,000 patients with UGIB, PUD was responsible for 27-40% of all bleeding episodes (Stabile, 2000). High-risk patient populations at risk for PUD include those with a history of alcohol abuse, chronic renal failure, and/or NSAID use. Duodenal ulcers are more common than gastric ulcers, but the incidence of bleeding is identical for both. In most cases, the bleeding is caused by the erosion of an artery in the base of the ulcer. Bleeding vessels larger than 1.5 mm in diameter are associated with an increased mortality rate. Initial endoscopic attempts to maintain hemostasis have a high failure rate. In approximately 80% of patients, bleeding from a peptic ulcer stops spontaneously (Stabile, 2000). A

10 minority of patients experience recurrent bleeding, and these cases are usually associated with risk factors for rebleeding. These factors include age older than 60 years, the presence of shock upon admission, coagulopathy, active pulsatile bleeding, and the presence of cardiovascular disease. These circumstances are associated with a poorer prognosis and a higher mortality rate. The appearance of the ulcer at the time of endoscopy provides important information regarding the risk of rebleeding. Forrest et al were the first to classify the stigmata of hemorrhage from peptic ulcers. Based on these classifications, the risk of recurrent bleeding can be predicted. The ulcers at highest risk for rebleeding are those that involve active arterial bleeding or those with a visible, protuberant, nonbleeding vessel in the base of the ulcer. The study not only correlated the incidence of rebleeding with the stigmata of recent bleeding and the endoscopic appearance of an ulcer, but also determined prognostic information regarding the need for surgery. Mortality was also correlated (Corson, 2001). Table 5. Ulcer Characteristics and Correlations Ulcer Characteristics Clean base Flat spot Adherent clot Visible vessel Active bleeding Prevalence Rebleeding Surgery Mortality Rate, % Rate, % Rate, % Rate, % 42 20 17 17 18 5 10 22 43 55 0.5 6 10 34 35 2 3 7 11 11

Upper GI endoscopy is the most effective diagnostic tool for PUD and has become the method of choice for controlling active ulcer hemorrhage. Failure of endoscopy to maintain hemostasis is one of the indications to initiate surgical intervention, especially in high-risk patients. A recent randomized prospective trial involved 92 patients with recurrent ulcer bleeding after initial endoscopic therapy for hemostasis who were randomized to surgery versus endoscopic re-treatment (Lau, 1999). The group randomized to another endoscopic attempt to control bleeding was found to have decreased transfusion requirements, 30-day mortality rates, and duration of ICU stay when compared to the surgical group. With the exception of a patient in shock who has a life-threatening recurrent hemorrhage, this study supports attempting another trial of endoscopy to control a bleeding ulcer. Regardless of the endoscopic therapy, 10-12% of patients with acute ulcerous hemorrhage require an operation as the definitive procedure to control the bleeding ulcer. In most circumstances, the operation is performed emergently, and the associated mortality rate is as high as 15-25%. Medical therapy used in conjunction with endoscopy involves proton pump inhibitor administration. Proton pump inhibitors decrease rebleeding rates in patients with bleeding ulcers associated with an overlying clot or visible nonbleeding vessel in the base of the ulcer. Consider transcatheter angiographic embolization in

11 patients who are poor surgical candidates. Because of the extensive collateral circulation of the upper GI tract, ischemic complications are rare. If 2 attempts at endoscopic control of the bleeding vessel are unsuccessful, avoid further attempts (ie, because of increased rebleeding and mortality rates) and pursue surgical intervention. The indications for surgery in patients with bleeding peptic ulcers are as follows: Severe life-threatening hemorrhage not responsive to resuscitative efforts Failure of medical therapy and endoscopic hemostasis with persistent recurrent bleeding A coexisting reason for surgery such as perforation, obstruction, or malignancy Prolonged bleeding with loss of 50% or more of the patient's blood volume A second hospitalization for peptic ulcer hemorrhage The operative treatment options for a bleeding duodenal ulcer historically include vagotomy, gastric resection, and drainage procedures. Each specific operative option is associated with its own incidence of ulcer recurrence, postgastrectomy syndrome, and mortality. When making an intraoperative judgment on how to best manage the bleeding ulcer, it is extremely important for the surgeon to be aware of these differences (Corson, 2001). Table 6. Recurrent Ulcer and Postgastrectomy Syndromes After Operations for Duodenal Ulcer Original Operation Proximal gastric vagotomy Truncal vagotomy and drainage Postgastrectomy Recurrence Mortality Syndrome Rate, Rate, % Rate, % % 10 7 5 20-30 30-50 0.1 <1 0-5

Truncal vagotomy and antrectomy 1 Billroth I or Billroth II Truncal vagotomy and antrectomy 5-10 Roux-en-Y

50-60

0-5

The 3 most common operations performed for a bleeding duodenal ulcer are as follows (Stabile, 2000): Truncal vagotomy and pyloroplasty with suture ligation of the bleeding ulcer Truncal vagotomy and antrectomy with resection or suture ligation of the bleeding ulcer Proximal (highly selective) gastric vagotomy with duodenostomy and suture ligation of the bleeding ulcer The purpose of the vagotomy is to divide the nerves to the acid-producing body and fundus of the stomach. This inhibits the acid production that occurs during the cephalic phase of gastric secretion. Although acid secretion is controlled, gastric motility and gastric emptying is affected, as indicated in the following table (Corson, 2001).

12 Table 7. Effects of Operations for PUD on Gastric Emptying and Motility Operation Proximal vagotomy gastric Antral Innervation Preserved Divided Liquid Emptying Fast Fast Fast Fast Solid Emptying Normal Slow Fast Fast

Truncal vagotomy

Truncal vagotomy Divided and drainage Truncal vagotomy Divided and antrectomy

Proximal vagotomy abolishes gastric receptive relaxation and impairs storage in the proximal stomach. As a result, a more rapid gastric emptying of liquids occurs. A drainage procedure is not required because the innervation of the antrum and pylorus is still intact. Because of this, the gastric emptying of solid food is not altered. The antropyloric mechanism still functions normally and continues to prevent duodenogastric reflux. In addition to having the same effects as a highly selective vagotomy in the proximal stomach, a truncal vagotomy also has marked effects on distal gastric motor function. It weakens distal gastric peristalsis, thus requiring the creation of a pyloroplasty to decrease the resistance to outflow from the stomach. Truncal vagotomy and suture ligation of a bleeding ulcer is a frequently used operation for treating UGIB in elderly patients with life-threatening hemorrhage and shock. The procedure can be performed rapidly, minimizing the time spent in the operating room under general anesthesia. The principles of suture ligation of a duodenal bleeding ulcer that involves the gastroduodenal artery require use of the 3-point ligation technique. The gastroduodenal artery is ligated both proximally and distally to the arterial bleeding site. The third suture is a horizontal mattress placed to control hemorrhage from the transverse pancreatic branch of the gastroduodenal artery. Failure to place this third stitch may result in recurrent bleeding that requires another emergent laparotomy of the abdomen. Vagotomy with antrectomy is reserved for patients whose conditions have failed to respond to more conservative attempts at surgical intervention and for those with aggressive and recurrent duodenal ulcer diathesis such as gastric outlet obstruction. When performing a highly selective vagotomy, the duodenostomy or the pyloroduodenostomy is closed anatomically, preserving the normal pyloric sphincter muscle. Most commonly, this operation is reserved for young, stable, low-risk patients. Although long-term follow-up care is still necessary, the recurrent ulcer rate is less than 10% at a mean follow-up of 3.5 years (Stabile, 2000). Much of what is now known about the operations performed for bleeding duodenal ulcers came from the era before the etiologic role for Helicobacter pylori

13 and NSAIDs in the development of peptic ulcers was understood. Reducing gastric acidity has been proven to be beneficial, with lower rebleeding rates when using high-dose omeprazole (Fallah, 2000). Although proton pump inhibitors seem to have an advantage, they have no affect on mortality. The diagnosis of H pylori infection is important in the management of patients with a complicated bleeding peptic ulcer. If a patient with a bleeding ulcer requires surgery, then knowledge of the patient's H pylori status becomes pertinent because it may help guide the decision to choose a particular surgical procedure, eg, simply oversewing the ulcer as opposed to performing an antiulcer operation. Many studies support the decision to manage the bleeding ulcer in conjunction with eradication of H pylori. In patients with H pylori infection, the rate of recurrent bleeding is extremely low. This is why documenting the presence of H pylori and aggressively treating the infection are important. Patients who are not infected with H pylori may require a subsequent acid-lowering surgical procedure or long-term medical therapy for recurrent ulcer disease and bleeding. Rates of H pylori infection are reportedly lower in patients with complicated ulcer disease compared to patients with uncomplicated ulcers. Hosking et al reported a 71% incidence of H pylori infection in patients with bleeding duodenal ulcers; patients with nonbleeding ulcers had an incidence of 93%. NSAIDs cause gastric and duodenal ulcers by inhibiting cyclooxygenase, which causes decreased mucosal prostaglandin synthesis and results in impaired mucosal defenses. Daily NSAID use causes an estimated 40-fold increase in gastric ulcer creation and an 8-fold increase in duodenal ulcer creation (Corson, 2001). Long-term NSAID use is associated with a 20% incidence in the development of mucosal ulceration (Corson, 2001). Medical therapy includes avoiding the ulcerogenic drug and beginning a histamine-2 (H2)receptor antagonist or a proton pump inhibitor that provides mucosal protection. H pylori is a gram-negative flagellated rod that plays a major etiologic role in the development of PUD. H pylori can be found in the antrum of the stomach in 95% of patients with a duodenal ulcer and in most patients with a gastric ulcer not associated with NSAID use. Eradication of H pylori can reduce the risk of rebleeding. Current antiH pylori regimens include a variety of drug combinations. Typically, an antimicrobial agent is combined with an H2-receptor antagonist or a proton pump inhibitor. The treatment regimens approved by the US Food and Drug Administration (FDA) have eradication rates for H pylori of 70-90% (Corson, 2001). Treatment regimens for H pylori infection (Corson, 2001) o Omeprazole: Administer 40 mg/d plus clarithromycin at 500 mg q8h for 2 weeks. Then administer omeprazole at 20 mg/d for 2 weeks. o Ranitidine bismuth citrate: Administer 400 mg q12h plus clarithromycin at 500 mg q8h for 2 weeks. Then administer ranitidine bismuth citrate at 400 mg q12h for 2 weeks. o Bismuth subsalicylate: Administer 525 mg q6h plus metronidazole at 250 mg q6h plus tetracycline at 500 mg q6h for 2 weeks plus an H2receptor antagonist for 4 weeks.

14 Lansoprazole: Administer 30 mg plus amoxicillin at 1 g plus clarithromycin at 500 mg q12h for 2 weeks. Poor compliance with the drug treatment can create antibiotic resistance and can fail to eradicate H pylori organisms. Diagnostic tests for H pylori can be divided into 2 groups, ie, (1) tests for patients who will undergo endoscopy and subsequently have a biopsy of the gastric mucosa performed and (2) tests for those who do not require endoscopy. The endoscopic tests include culture, histology, and a rapid urease test (RUT). RUTs are based on the urease-producing activity of H pylori. If urease is present, then urea is hydrolyzed, thus releasing ammonia that raises the pH. The CLO test and the HP test are 2 of the tests now commercially available. Available nonendoscopic tests are the urea breath test and serologic testing. The urea breath test and the endoscopic biopsy results rely on the bacterial load to detect H pylori; therefore, the use of proton pump inhibitors and H2-receptor antagonists should be stopped 2-4 weeks before performing these studies. The RUT has an excellent specificity, approaching 100%, although it has a false-negative rate of approximately 10% (Laine, 1999). Therefore, performing a biopsy is important to document the presence of H pylori or inflammation with a cellular infiltrate (neutrophils). Bleeding gastric ulcer The surgical management of bleeding gastric ulcers is slightly different from that of duodenal ulcers, but the concepts are identical. The 3 most common complications of a gastric ulcer that mandate emergent surgical intervention are hemorrhage, perforation, and obstruction. The goals of surgery are to correct the underlying emergent problem, prevent recurrent bleeding or ulceration, and exclude malignancy. A bleeding gastric ulcer is most commonly managed by a distal gastrectomy that includes the ulcer with a gastroduodenostomy or a gastrojejunostomy reconstruction. The common operations for the management of a bleeding gastric ulcer include (1) truncal vagotomy and pyloroplasty with a wedge resection of the ulcer, (2) antrectomy with wedge excision of the proximal ulcer, (3) distal gastrectomy to include the ulcer with or without truncal vagotomy, and (4) wedge resection of the ulcer only. The choice of operation depends on the location of the ulcer and the hemodynamic stability of the patient to withstand an operation. Five types of gastric ulcers occur, based on their location and acid-secretory status. Type 1 gastric ulcers are located on the lesser curvature of the stomach at or near the incisura angularis. These ulcers are not associated with a hypersecretory acid state. Type 2 ulcers represent a combination of 2 ulcers that are associated with a hypersecretory acid state. The ulcer locations occur in the body of the stomach in the region of the incisura. The second ulcer occurs in the duodenum. Type 3 ulcers are prepyloric ulcers. They are associated with high acid output and are usually within 3 cm of the pylorus. Type 4 ulcers are located high on the lesser curvature of the stomach and (as with type 1 ulcers) are not associated with high acid output.
o

15 Type 5 ulcers are related to the ingestion of NSAIDs or aspirin. These ulcers can occur anywhere in the stomach. A vagotomy is added to manage type 2 or type 3 gastric ulcers. These ulcers arise in the pyloric channel or the prepyloric area and are associated with acid hypersecretion physiology. Patients who are hemodynamically stable with intermittent bleeding requiring blood transfusions should undergo a truncal vagotomy and distal gastric resection to include the ulcer for type 1, 2, and 3 ulcers. In patients who present with life-threatening hemorrhage and a type 1, 2, or 3 ulcer, biopsy and oversew or excision of the ulcer in combination with a truncal vagotomy and a drainage procedure should be considered. Patients with type 4 ulcers usually present with hemorrhage. The left gastric artery should be ligated, and a biopsy should be performed on the ulcer. Then, the ulcer should be oversewn through a high gastrotomy. Rebleeding rates for the procedures that keep the ulcer in situ range from 20-40% (Corson, 2001). Gastric bleeding in the immediate postoperative period from recurrent PUD is initially best managed by endoscopic or angiographic means. If reoperation is required, gastric resection is usually indicated because a repeat vagotomy is not reliable. A more definitive solution is warranted. Stress gastritis Acute stress gastritis is a disease process characterized by diffuse superficial mucosal erosions that appear as discrete areas of erythema (Corson, 2001). The bleeding is usually mild and self-limiting and rarely progresses to lifethreatening hemorrhage. In ICU patients, the incidence of clinically significant GI bleeding (eg, hypotension, transfusion) from this process was 1.5% (Stollman, 2004). Stress gastritis and mucosal ulceration are historically associated with (1) head injuries with associated elevations in intracranial pressure and (2) burn injuries. These stress ulcers are called Cushing ulcer and Curling ulcer, respectively. Predisposing clinical conditions have the potential to alter local mucosal protective barriers such as mucus, bicarbonate, blood flow, and prostaglandin synthesis. Any disease process that disrupts the balance of these factors results in diffuse gastric mucosal erosions. This is most commonly observed in patients who have undergone episodes of shock, multiple trauma, acute respiratory distress syndrome, systemic respiratory distress syndrome, acute renal failure, and sepsis. The principal mechanisms involved are decreased splanchnic mucosal blood flow and altered gastric luminal acidity. Knowledge of the predisposing conditions for stress ulceration allows the clinician to identify patients at risk for developing gastritis and GI bleeding. Treatment in this group of high-risk patients should focus on prevention. This is best accomplished by treating the underlying causes of ulceration. Aggressive support of hemodynamic parameters ensures adequate mucosal blood flow. In addition, several strategies have evolved to treat gastric luminal acidity. Histamine receptor antagonists (HRAs) have proven to be the most effective at controlling stomach pH. Proton pump inhibitors (PPIs) are superior to the HRAs at suppressing acid; however, their role in stress ulceration prophylaxis is still being studied (Stollman, 2004).

16 Stress-related bleeding usually occurs 7-10 days after the initial insult but may manifest sooner. Initially, endoscopy is the most important diagnostic tool. The acute superficial erosions are multiple, begin in the fundus, and progress toward the antrum. Ninety percent of patients stop bleeding with conservative medical therapy that includes NGT lavage and gastric acid controlling medications to maintain the gastric luminal pH above 5.0 (Cameron, 1995). Endoscopic hemostasis is attempted using electrocoagulation, laser, or injection therapy. Selective angiographic catheterization of the left gastric artery may be attempted with selective infusion of vasopressin (48-72 h) or embolization using Gelfoam, coils, or autologous clot to embolize the left gastric artery. Regardless of the angiographic technique used, it is often unsuccessful because of the rich and extensive submucosal plexus and collateral circulation within the stomach. Surgical intervention becomes necessary if nonoperative therapy fails and blood loss continues. The goals of operative treatment are to control bleeding and to reduce recurrent bleeding and mortality. These patients are at extremely high risk, and the most expeditious procedure is the best option. Simply oversewing an actively bleeding erosion is sometimes effective enough to control the bleeding. In the setting of life-threatening hemorrhage not amenable to endoscopic control, gastric resection with or without vagotomy with reconstruction may be necessary. The type of gastric resection depends on the location of the gastric erosions, ie, whether they are proximal or distal. The options are antrectomy and subtotal, near total, or total gastrectomy. Operative mortality rates range from 30-100% (Cameron, 1995). The choice of the initial operation must be made with an understanding of the patient's condition, the amount and location of gastric disease, and an accurate assessment of one's technical ability to rapidly and safely perform a gastric resection. Managing the underlying insult causing the gastric stress ulcerations is also important. This involves supportive measures to maintain acceptable hemodynamic parameters, to provide adequate nutritional support in the critically ill patient, and to treat sepsis if present. Portal hypertensionrelated bleeding Variceal bleeding is one of the most alarming life-threatening complications of cirrhosis. Sixty percent of patients with cirrhosis develop esophageal varices. Thirty percent of these patients bleed from their varices within 2 years of their diagnosis, and 50% bleed at some point during their lifetime. The mortality rate for variceal bleeding is 30-50%, which is much higher than any other cause of UGIB (Corson, 2001). Patients with portal hypertension but normal liver function (eg, portal vein thrombosis, idiopathic portal hypertension, schistosomiasis) have better outcome and survival rates compared to those with liver dysfunction and other organ failure. The presence of portal vein pressure greater than 10 mm Hg defines portal hypertension. Normal portal venous pressure ranges from 5-10 mm Hg, which provides a portal blood flow rate of approximately 1 L/min through the hepatic sinusoids. Portal hypertension can be classified according to the anatomic location within the portal system that is the site for increased resistance to portal

17 flow. The 3 classifications are presinusoidal, sinusoidal, and postsinusoidal. This organized classification is clinically useful to differentiate patients with hepatocellular dysfunction from those with normal liver function, as noted in the following table (Corson, 2001). Table 8. Causes and Sites of Block for Portal Hypertension Presinusoidal Extrahepatic Congenital Umbilical sepsis Trauma Hypercoagulation state Malignant occlusion Intrahepatic Schistosomiasis Congenital hepatic Fibrosis Primary biliary Cirrhosis Patients with presinusoidal portal hypertension have normal liver function; patients with sinusoidal causes for portal hypertension have cirrhosis and destroyed sinusoidal anatomy. Regardless of the cause of the increased portal pressure, the resistance to portal blood flow leads to the development of 4 wellrecognized collateral vascular systems, ie, (1) the esophageal submucosal venous plexus, (2) the cardiac vein of the stomach, (3) the retroperitonealumbilical system, and (4) the hemorrhoidal system. The bleeding risk for esophageal varices becomes significant once the intravariceal pressure reaches a threshold level. The increased tension (T) within the varices is directly related to its intravascular pressure (P) and radius (R) and is inversely related to its wall thickness (W), ie, T = P X R / W. The life-threatening bleeding that occurs is the result of the high intravascular tension. Of the patients newly diagnosed with cirrhosis each year, 30% have compensated disease; however, it is the 60% with uncompensated disease who have varices (Corson, 2001). Estimates from prospective studies indicate that the overall incidence of esophageal varices in patients with cirrhosis is 8% each year (Vargas, 1999). The clinical implication of this is extremely important because all patients newly diagnosed with cirrhosis should undergo a screening endoscopy. Sinusoidal Cirrhosis Postsinusoidal Hepatic veins Budd-Chiari syndrome Venoocclusive disease

18 Because the mortality rate associated with an acute variceal bleed is extremely high, varices at high risk for bleeding must be recognized. An Italian prospective study analyzed the endoscopic and clinical features of cirrhosis and predicted the likelihood of variceal bleeding based on these parameters (North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices, 1998). The study revealed that the risk of the first variceal bleed may be predicted based on variceal size, the presence of red wales, and a Child-Pugh score greater than or equal to 8 points (Corson, 2001). Table 9. Child-Pugh Classification Parameter Bilirubin Albumin Increase in PT Ascites Encephalopathy 1 Point 2 Points 3 Points <2 >3.5 1-3 None None 2-3 2.8-3.5 4-6 Slight 1-2 >3 <2.8 >6 Moderate 3-4

In addition to the Child-Pugh classification system, the model for end-stage liver disease (MELD) score has been used to predict outcomes in this patient population. The MELD score uses 3 objective measurements: creatinine, bilirubin, and INR to quantify the degree of hepatic dysfunction. It is currently used for allocating patients to liver transplant waiting lists (Amitrano, 2005). In regard to patients with active variceal bleeding, the MELD score can be used to identify short-term mortality risk (Amitrano, 2005). This information can then be used to direct a more intense surveillance and therapeutic strategy in these high-risk patients. The MELD score calculation is as follows (Kamath, 2001): (0.957 x log(e) (creatinine mg/dl)+0.378 x log(e) (bilirubin mg/dl) +1.120 x log(e) (INR) +0.643)x10 Other studies have studied the correlation between the portal-hepatic vein gradient (PHVG) and the risk for bleeding. The pressures are measured in the hepatic vein when the vein is both occluded and in the free position. This is performed using the same principles as those for measuring pulmonary artery pressures and pulmonary artery occlusion pressures. The occluded hepatic vein pressure is an indirect measurement of the hepatic sinusoidal pressure. The difference between the occluded and free hepatic vein pressure is called the PHVG. Garcia-Tsao et al observed that all the patients in their study bled when the PHVG was greater than 12 mm Hg. Once an initial acute bleed occurs, early rebleeding ensues in 20-50% of patients within the first 7-10 days. Within approximately the first 6 weeks after the initial bleed, the risk for rebleeding drops to what it was before the initial bleed. The risk for rebleeding within the first year is 30%. The initial management of a patient with an acute variceal bleed involves aggressive efforts at resuscitation as previously outlined. Admit the patient to an

19 ICU setting in which a multidisciplinary team approach can be used to treat the patient. Follow the ABCs, but avoid overhydrating the patient because volume overload can increase portal hypertension and provoke more bleeding. Since Gilbert and Vailaret coined the term portal hypertension in 1906, many therapeutic modalities have been developed to manage acute variceal hemorrhage and subsequent elective treatments for prevention of recurrent bleeding. These treatment strategies can be classified into (1) pharmacologic therapy, (2) endoscopic therapy, (3) tamponade, (4) decompressive therapy (radiologic and surgical), and (5) liver transplantation. Pharmacologic therapy The patient with acute variceal bleeding may initially be treated with intravenous vasopressin and nitroglycerine, somatostatin, or one of its analogs (eg, octreotide). Vasopressin is a potent splanchnic and systemic vasoconstrictor, including coronary vasoconstriction. Nitroglycerin should be concomitantly administered to titrate and maintain the SBP in the range of 90-100 mm Hg. Nitroglycerin should be initiated at 40 mcg/min to protect the coronary arteries from the profound adverse cardiovascular effects of the vasopressin. The intravenous infusion of vasopressin is started at 0.2-0.4 U/min. A meta-analysis from 1995 reviewed vasopressin versus untreated controls and revealed that vasopressin was clearly superior in arresting hemorrhage but showed no survival benefit (D'Amico, 1995). In Europe, a newly developed prodrug called terlipressin has been used that has advantages over vasopressin (Vargas, 1999). Terlipressin has a longer halflife with a biphasic vasoconstriction profile. The drug first has systemic vascular effects that are then steadily converted into a more effective vasoconstriction of the splanchnic bed. Randomized control trials comparing terlipressin with placebo have shown clear benefit in terms of bleeding control and survival (Patch, 1997). This drug is not yet available in the United States. Recombinant coagulation factor VII is a synthetic coagulation factor that is currently used to treat acute bleeding episodes in patients with hemophilia. It has recently been trialed in randomized studies as an adjunct to standard methods of controlling variceal hemorrhage. Patients in end-stage liver disease lose the synthetic capacity to produce these coagulation factors, most notably factor VII (Bosch, 2005). As a result, it has been postulated that replacing factor VII in cirrhotics may aid in controlling acute variceal hemorrhage. A recent double-blind trial has shown increased success at controlling bleeding endpoints in more severe cirrhotics (classes B and C) treated with factor VII compared to placebo (Bosch, 2005). Although results are promising, cost-effective dosing and timing of therapy are yet to be determined (Jimenez-Saenz, 2005). Endoscopy in variceal bleeding Since rigid endoscopy was first introduced in 1939 for the treatment of variceal bleeding, its evolution into flexible endoscopy has enabled it to become the cornerstone of current management for acute variceal hemorrhage. The 2 main endoscopic techniques available to control variceal bleeding are endoscopic sclerotherapy and endoscopic variceal band ligation. Endoscopic sclerotherapy involves injecting a sclerosing agent, such as ethanolamine or polidocanol, into the varix lumen (intravariceal) or immediately adjacent to the vessel (paravariceal)

20 to create fibrosis in the mucosa overlying the varix, which leads to hemostasis. Endoscopic variceal banding ligation consists of the placement of a rubber band around the varix. This technique is performed by first sucking the varix into a sheath attached to the distal end of the endoscope. Once the varix is suctioned into the sheath, a trigger device allows the deployment of a rubber band around the varix, a procedure that strangulates the varix. Each varix is usually ligated as many times as necessary to make it no longer visible. The rubber bands slough off in the following 24-72 hours, leaving a shallow ulcer behind. Ulceration is less severe than with sclerotherapy. A recent meta-analysis of all the randomized controlled trials comparing sclerotherapy with variceal banding found significantly lower mortality rates, variceal rebleeding, esophageal perforation, and stricture formation with variceal banding therapy (Laine, 1995). In addition, evidence accumulating in the literature documents the safety of endoscopic banding in prophylactic therapy for patients at high risk for esophageal variceal hemorrhage (Lay, 1997). This benefit has not been established with endoscopic sclerotherapy. The single potential downside to endoscopic variceal banding is poor visualization of the varices when massive bleeding is present. This is due to the 30% reduction in the field of vision through the endoscope caused by the plastic sheath used to deploy the rubber band (Vargas, 1999). The drawback to sclerotherapy is its potential for life-threatening complications such as perforation, ulceration, and stricture of the esophagus. The initial success in arresting hemorrhage comparing the 2 techniques is comparableas high as 95% in some studies (Vargas, 1999). Variceal rebleeding rates are lower with band ligation compared to sclerotherapy, which has rebleeding rates as high as 50% (D'Amico, 1995). In summary, endoscopic variceal banding is a more effective technique resulting in faster variceal eradication and fewer complications than endoscopic sclerotherapy. Endoscopic surveillance at 3-month intervals during the first year is important to detect recurrent and new varices when they are small and have a low risk of bleeding (Corson, 2001). The other available endoscopic option is the use of tissue adhesives, which are useful for both esophageal and gastric varices. Native cyanoacrylate is a liquid tissue adhesive used frequently in Europe. It has a consistency similar to water, which makes it easy to manipulate down the endoscope. After injecting the substance into the varix, the blood mixes with the adhesive agent and rapidly polymerizes into a hard glue. The cyanoacrylate then plugs the lumen of the varix and creates hemostasis. Several weeks (ie, 2 wk to 3 mo) after the initial injection, the overlying mucosa sloughs off and the adhesive cast is passed through the GI tract. Cyanoacrylate is 90% successful in achieving hemostasis in patients with acute bleeding from either gastric or esophageal varices (D'Imperio, 1996). The early rebleeding rate is 0-28%. The risks of using these adhesive agents are the potential systemic effects and permanent damage to the endoscope from exposure to the adhesives at the lens of the scope (Sarin, 2001). The potential systemic effects are related to the risk of thrombotic complications. Cerebral stroke from anomalous right-to-left shunts, fatal

21 pulmonary embolization, portal vein embolization, splenic infarction, and retrogastric abscess have all been reported in the literature (Sarin, 2001). Oho et al performed a prospective randomized trial involving 53 patients that compared cyanoacrylate glue to sclerotherapy in patients with acute gastric variceal bleeding. Cyanoacrylate was more effective in achieving hemostasis (93% vs 67%). The need for emergent surgical intervention was also reduced in patients initially injected with the glue tissue adhesive. Other agents used include bovine fibrin or fibrin glue. Polidocanol, a 2-component fibrin glue, has shown promise as an effective agent with a decreased incidence in mucosal damage or posttherapy ulceration that is characteristic of sclerosants and tissue adhesives (Zimmer, 1998). Bovine thrombin used as an injectable hemostatic agent is not yet approved by the FDA for use in the United States. Balloon tamponade Since the 1990s, the use of balloon tamponade for the emergent control of esophageal variceal bleeding has steadily decreased and is now usually indicated for only the 10% of patients whose acute bleeding episode is not controlled endoscopically (Corson, 2001). The use of these tubes can be a life-saving maneuver when medical and endoscopic efforts fail to control the bleeding. Although temporary control of the bleeding is effective in 85% of cases, recurrent bleeding with release of the tamponade occurs in most patients (Kupfer, 2000). In addition to their limited effectiveness, the tubes are associated with a 20% complication rate that includes airway obstruction, aspiration, and esophageal necrosis with rupture (Kupfer, 2000). Because of the severe life-threatening complications and limited use, the tubes are used only as a temporary measure while the patient is resuscitated. The tubes act as a bridge to help stabilize the patient until a time when the patient is prepared for either a repeat endoscopy procedure or a portal pressure decompression through a radiological or surgical method. The 2 most commonly used tubes are the Sengstaken-Blakemore tube and the Minnesota tube. These tubes have an esophageal balloon and a gastric balloon that are inflated to produce a tamponade effect after confirming appropriate anatomical placement. When deciding to employ this line of therapy, seriously consider endotracheal intubation to secure and protect the patient's airway. Although balloon tamponade effectively controls acute variceal hemorrhage, a meta-analysis of studies comparing balloon tamponade to drug therapy and endoscopic sclerotherapy revealed no survival advantage in the balloon group. The physician must be familiar with the steps involved in placement of the tubes to avoid the unnecessary complications (Kupfer, 2000). Complications associated with balloon tamponade are as follows: Major complications o Esophageal rupture o Tracheal rupture o Duodenal rupture o Respiratory tract obstruction o Aspiration o Hemoptysis

22 Tracheoesophageal fistula o Jejunal rupture o Thoracic lymph duct obstruction o Esophageal necrosis o Esophageal ulcer Minor complications o Nasopharyngeal bleeding o Chest pain o Balloon impaction and/or migration (nausea and vomiting) o Alar necrosis The tube is first introduced into the stomach, and a small amount of air is injected into the gastric balloon. A radiograph is then obtained to confirm placement in the stomach. Once proper placement is confirmed, the gastric balloon is inflated with 300-350 mL of air and is pulled up into the gastric fundus, compressing the gastroesophageal junction. The tube is secured to the facemask of a football helmet placed on the patient's head. The esophageal balloon is then inflated to a pressure of 40 mm Hg. Another radiograph is obtained to confirm proper placement of both tubes. Deflate the esophageal balloon every 4 hours for 15 minutes to avoid esophageal pressure necrosis. Do not leave the entire tube in place for more than 24-48 hours. Transjugular intrahepatic portosystemic shunt Transjugular intrahepatic portosystemic shunt (TIPS) decompression of the portal system can be achieved through either radiologic or surgical methods. The goal is to reduce intravariceal pressure to less than 12 mm Hg. The TIPS procedure has become the most frequently employed method; because of its effectiveness, it is considered the standard of therapy for bleeding esophagogastric varices that are unresponsive to endoscopic and pharmacologic first-line treatment. The placement of a TIPS reduces the outflow hepatic resistance, lowers portal pressure, and diverts portal blood flow from gastroesophageal collaterals through the stent. The procedure takes approximately 1-2 hours to perform. The right internal jugular vein is cannulated, and a coaxial catheter system is used to maintain access to either the middle or right hepatic vein. A catheter is passed over a wire and advanced through the hepatic parenchyma to obtain access to the portal system. This track created between the systemic venous system and the portal system is balloon-dilated so that a self-expanding metal stent can be inserted and deployed in the track. The wall stent is dilated to approximately 10-12 mm in diameter. Once the stent is in proper position, pressures are measured in the portal vein, the stent, and the right atrium. The goal of the procedure is to reduce the portal-to-atrial pressure gradient to less than 12 mm Hg. Another measurement used is the PHVG. Measurement of the free hepatic and wedged hepatic pressures before and after the procedure should document a decrease of the PHVG to less than 12 mm Hg. The TIPS procedure controls variceal bleeding in more than 90% of patients (Sarin, 2001). The rebleeding rate is 16-30% at 1-year follow-up, and this is most commonly related to stenosis of the intrahepatic shunt or obstruction of the stent
o

23 (Sarin, 2001). Shunt dysfunction occurs in approximately 50-60% of patients at 6 months (Sarin, 2001). Close observation and repeat interventions are required to maintain shunt patency. Careful monitoring with Doppler ultrasound imaging at regular monthly intervals has led to better control of bleeding. Redilation of the stent is performed when the flow pattern through the shunt changes. This has increased the 1-year patency rate to 83-85% (Corson, 2001). The risk of inducing hepatic encephalopathy is 25-35%, but this can usually be managed with protein restriction and lactulose (Sanyal, 1994). The 30-day mortality rate is 14-16%, with most deaths occurring in patients with Child class C cirrhosis as a result of multisystem organ failure. The TIPS procedure is usually considered in patients who have not responded to first-line treatment using pharmacologic and endoscopic therapy. It may also be necessary in patients who required balloon tamponade and have recurrent bleeding refractory to repeat endoscopic attempts at hemostasis. The severity of rebleeding and the rate of progression of the underlying liver disease are major factors that should be considered when determining the appropriate timing for portal decompression and the method to accomplish the task. For patients who do not respond early and who have recurrent bleeding after esophageal variceal banding, rapid decompression is indicated. In patients who have minor rebleeding after variceal obliteration by endoscopic techniques, especially those who have not bled until 1-2 years later, a repeat endoscopic banding procedure is a more reasonable approach. The need for liver transplantation should also be considered during the initial evaluation of patients with UGIB and portal hypertension. For patients with Child class A or B cirrhosis with preserved liver function, portal decompression is preferable to transplantation. For patients with Child class C cirrhosis, the TIPS procedure can be used as a temporizing measure to provide a bridge until transplantation can be arranged. Surgical shunts Surgery for bleeding esophagogastric varices is the most reliable method to control acute hemorrhage and is associated with recurrent bleeding rates of less than 10% (Corson, 2001). The goal of a surgical shunt is to effectively decrease the portal venous hypertension and its adverse effects without compromising liver function. Because of the effectiveness and popularity of the available endoscopic therapies and the emergence of the TIPS procedure in the 1990s, the role of surgical decompression of the portal system is reserved for patients in whom initial nonsurgical management has failed and who still have preserved liver function. The widespread use of endoscopic techniques and the introduction of the TIPS procedure have made surgery a less attractive choice for acute and chronic variceal bleeding. The operative approaches for bleeding esophagogastric varices consist of 2 general concepts: (1) decompression of the high-pressure portal venous system into a low-pressure systemic venous system and (2) devascularization of the distal esophagus and proximal stomach (Stabile, 2000). The following options are currently available to the surgeon:

24 Portacaval shunt (end-to-side) Portocaval shunt (side-to-side) Small-diameter interposition graft Mesocaval shunt Large-diameter interposition graft Small-diameter interposition graft Distal splenorenal (ie, Warren) shunt Esophagogastric devascularization, esophageal transaction, and reanastomosis Orthotopic liver transplantation Splenectomy (for splenic vein thrombosis) Practitioners may well ask whether one shunt is superior to the others and whether one shunt controls bleeding better than the others. Each type of shunt has a bleeding control rate greater than 90% (Corson, 2001). The difference between shunts is the incidence of encephalopathy and the risk of worsening ascites. Depending on whether the shunt completely decompresses the portal venous system, portosystemic shunts can be classified as nonselective, selective, or partial. The goal of selective or partial shunts, in addition to decreasing portal venous hypertension, is to preserve portal blood flow to the liver in order to avoid compromising liver perfusion and function. Data from many randomized control trials that have compared the various types of shunts indicate that encephalopathy occurs in 10-15% of patients after a selective shunt (distal splenorenal), in 10-20% after a partial shunt, and in 30-40% after a total shunt (Corson, 2001). The differences in survival between all the different shunts are not significantly different, with mortality rates of approximately 5% (Corson, 2001). Table 10. Surgical Shunts for Treatment of Variceal Hemorrhage

Procedure Nonselective shunts End-to-end portocaval shunt Side-to-side portocaval shunt Selective shunts Distal splenorenal shunt Partial shunts Polytetrafluoroethylene interposition graft portocaval/mesocaval

Advantage

Disadvantage

Low bleeding rate Low bleeding rate

Encephalopathy Liver hypoperfusion

Few develop encephalopathy

Technically more difficult

Maintain liver perfusion; lower encephalopathy

Graft stenosis or thrombosis

25 shunt rate; low bleeding rate

Nonselective shunts completely divert portal blood flow from the liver. Because the patent shunt has lower resistance to blood flow than the cirrhotic liver, the portal blood flow through the hepatic vein is reversed. None of the nonselective shunts maintains hepatic portal perfusion unless they become stenotic. Although these shunts are the most effective at alleviating ascites, their incidence of encephalopathy is 25-50% because of the complete diversion of portal blood flow from the liver (Corson, 2001). The current indications for nonselective shunting are (1) in the emergency setting to control bleeding; (2) as a long-term bridge to liver transplantation in patients in whom endoscopic, pharmacologic, and TIPS therapies have failed; and (3) in patients who are bleeding from colonic and stomal varices. For patients in whom liver transplantation is anticipated, avoid dissection in the porta hepatis when creating a shunt. In this situation, an interposition mesocaval shunt is a better choice for a nonselective shunt. Selective shunting decompresses the varices while maintaining hepatopetal blood flow in the remainder of the portal system. The distal splenorenal (Warren) shunt (DSRS) is the most commonly used selective shunt. DSRS placement is performed by anastomosing the distal end of the splenic vein to the side of the left renal vein. The portal venous system is thereby divided into a decompressed gastrosplenic venous circulation and a high-pressure superior mesenteric venous circulation that maintains hepatopetal blood flow to the liver. The liver maintains portal perfusion, and the incidence of encephalopathy is lower (10-15%) (Corson, 2001). Worsening of ascites can occur because of the dissecting and dividing of lymphatics near and around the left renal vein. Avoid this shunt in patients with medically intractable ascites. The splenic vein should be 6-7 mm in diameter to avoid a high risk of shunt thrombosis. Selective shunts are best performed in patients with nonalcoholic portal hypertension. This is because patients with alcoholic cirrhosis tend to develop collaterals around the shunt, resulting in less portal flow in 50% of patients at 1-year follow-up (Corson, 2001). The objective of using partial shunts is the same as that of selective shunts, ie, to decompress varices while maintaining hepatic portal perfusion. This type of shunt has been popularized by Sarfeh and colleagues over the last 10 years. Sarfeh et al described using a small-diameter, polytetrafluoroethylene, interposition graft portocaval shunt combined with ligation of the coronary vein (Sarfeh, 1983). When the graft diameter is less than or equal to 10 mm, portal perfusion is maintained. Sarfeh et al have shown maintenance of portal perfusion in more than 80% of patients with 8-mm grafts. Shunt stenosis occurs in 15-20% of patients, and thrombosed grafts can be opened and dilated by interventional radiographic techniques. The incidence of encephalopathy is approximately 15%. One of the major controversies in the surgical management of bleeding varices has been the role of devascularization procedures. In 1973, Sugiura and Futagawa introduced their experience with a surgical technique that they developed involving gastroesophageal devascularization. The principle of the

26 Sugiura procedure is to divide the esophageal and gastric venous plexus from the portal system, while intentionally preserving the extra esophageal systemic venous collaterals to the azygous system. The left gastric vein must be preserved during the gastric devascularization to allow drainage of the systemic venous system into the azygous vein. Initially, this was a 2-stage procedure that first involved devascularizing the upper stomach and esophagus up to the left pulmonary vein through a thoracotomy, followed by a laparotomy a few weeks later. The initial Japanese series reported a low mortality rate (4-12%), with effectiveness in the prevention of recurrent bleeding (1.5-16%) (Futagawa, 1979). This procedure has not been as successful in the Western Hemisphere. The Western world has not been able to reproduce Sugiura's exceptional results. One of the largest series published in the literature was by Orozco et al from Mexico City, which involved 100 patients. In this series, the overall mortality rate was 22% but reached as high as 80% in the subgroup of patients with Child class C cirrhosis. In the subgroup of patients with Child class A and B cirrhosis, the postoperative mortality rate was 12% and 31%, respectively. The rebleeding rate was 10%, and the incidence of postoperative encephalopathy was 3%. In patients undergoing the procedure electively versus emergently, the mortality rates were more than 11% versus 41%. Several possible explanations may account for the differences between the Japanese results and those observed in the Western Hemisphere (Selzner, 2001). For example, Sugiura had a low percentage of Child class C patients; additionally, alcoholism is a rare cause of portal hypertension in Japan compared to Western experiences. Patients with alcohol abuse and alcoholic cirrhosis are at higher risk because of concomitant comorbid illnesses and malnutrition. Lastly, Sugiura performed the procedure in an elective setting. The procedure was primarily performed for prophylaxis of variceal bleeding. In 2001, an article by Selzner described the technique of the modified Sugiura procedure, which involves a 1-stage transabdominal approach. A splenectomy is first performed. Next, the stomach is devascularized to include the upper two thirds of the lesser and greater curvatures. The devascularization of the esophagus is carried up 6-7 cm through the diaphragmatic hiatus. The left gastric vein is preserved during this maneuver. Next, a circular stapler is introduced 4-6 cm into the esophagus through an anterior gastrotomy. The circular stapler is fired, allowing concomitant transection and reanastomosis of the esophagus. A pyloroplasty is then performed after reinforcing the esophageal anastomosis. Preservation of the anterior and posterior vagus and Nissen fundoplication have been described by other authors as part of the procedure. In this study, the modified Sugiura procedure was offered only to patients who had Child class A or B cirrhosis with well-preserved synthetic liver function and contraindications to DSRS, TIPS, and liver transplantation. Patients with Child class C cirrhosis and those with Child class B cirrhosis with ascites were excluded from the procedure. Through appropriate patient selection, the mortality rate in this study was 9% and the rebleeding rate was 7%. These are satisfactory numbers considering the high-risk patient population and the dismal outcomes after exhausting other therapeutic alternatives. Patient selection is extremely

27 important when considering surgical intervention. Preserved liver function and an elective setting are key factors in predicting morbidity and mortality. An algorithm has been developed for patients with recurrent esophageal variceal bleeding in whom endoscopic therapy has failed (Selzner, 2001). Patients with marginal liver function (ie, Child class B or C cirrhosis) should receive TIPS as a bridge to liver transplantation. Patients with preserved liver function (ie, Child class A or B cirrhosis without ascites) should first be considered for a DSRS. If the patient is not a candidate for a surgical (ie, DSRS) shunt because of a previous splenectomy, small splenic vein, or splenic vein thrombosis, then the modified Sugiura procedure should be considered (Selzner, 2001). The Sugiura procedure is indicated when other therapeutic modalities are not possible or have failed. Mallory-Weiss syndrome Mallory-Weiss tears account for 15% of acute upper GI hemorrhage (Stabile, 2000). Kenneth Mallory and Soma Weiss first described the syndrome in 1929 (Cameron, 1995). The massive UGIB results from a tear in the mucosa of the gastric cardia. This linear mucosal laceration is the result of forceful vomiting, retching, coughing, or straining. These actions create a rapid increase in the gradient between intragastric and intrathoracic pressures, leading to a gastric mucosal tear from the forceful distention of the gastroesophageal junction (Cameron, 1995). In 80-90% of cases, this is a single 1.75- to 2.5-cm mucosal tear along the lesser curve of the stomach just distal to the gastroesophageal junction (Cameron, 1995). The initial step in management is to first make the correct diagnosis. Distinguishing Mallory-Weiss syndrome from Boerhaave syndrome is critical. Although both entities share a common pathogenesis, their management is completely different. Boerhaave syndrome represents a full-thickness transmural laceration with perforation of the esophagus. A Gastrografin swallow helps confirm the presence of the perforation in most cases, and prompt surgical intervention is necessary to prevent mediastinitis and sepsis. On the other hand, surgical intervention in Mallory-Weiss syndrome is required to achieve hemostasis in only 10% of cases (Cameron, 1995). The bleeding from a Mallory-Weiss tear spontaneously ceases in 50-80% of patients by the time endoscopy is performed (Cameron, 1995). For other patients in whom bleeding is visualized at endoscopy, the endoscopic treatment options are electrocoagulation, heater-probe application, or sclerotherapy. In a recent series published by Bataller et al, hemostasis was achieved in 100% of patients with Mallory-Weiss tears by using endoscopic sclerotherapy with epinephrine (1:10,000) and 1% polidocanol. Other nonoperative therapies are reserved for when the endoscopic attempts at creating hemostasis have failed. Other available options are angiographic intra-arterial infusion of vasopressin or transcatheter embolization of branches of the left gastric artery using Gelfoam. Avoid the balloon tamponade technique using the Sengstaken-Blakemore tube in this particular circumstance because this apparatus may extend the mucosal laceration into a transmural laceration with perforation (Cameron, 1995). Surgical intervention is indicated in patients with continued bleeding after failed attempts at nonoperative therapies. Bleeding from the gastroesophageal

28 junction is visualized through an anterior gastrotomy. Once the tear is localized, the bleeding is controlled by oversewing the lesion. The overall mortality rates of patients who require emergent surgery is 15-25%, in contrast to a mortality rate of 3% or less for patients whose bleeding stops by the time of the initial endoscopy (Cameron, 1995). Dieulafoy lesion Dieulafoy lesion, first described in 1896, is a vascular malformation of the proximal stomach, usually within 6 cm of the gastroesophageal junction along the lesser curvature of the stomach. However, it can occur anywhere along the GI tract. This lesion accounts for 2-5% of acute UGIB episodes (Larson, 1986). Endoscopically, the lesion appears as a large submucosal vessel that has become ulcerated. Because of the large size of the vessel, bleeding can be massive and brisk. The pathogenesis of the vessel rupture is usually in the setting of chronic gastritis, which may induce necrosis of the vessel wall. Alcohol consumption is reportedly associated with the Dieulafoy lesion. In a review of 149 cases, the Dieulafoy lesion mostly occurred in men and mostly in those in their third to tenth decade (Reilly, 1991). The initial endoscopic management of this lesion can be highly successful. In a report by Norton et al describing their experience with 90 Dieulafoy lesions, endoscopic management achieved primary hemostasis in 96% of cases. Contact thermal ablation with a heater probe is the most effective technique, with or without the combined use of epinephrine to slow or stop the bleeding prior to applying the heater probe. No studies have been performed that compare surgical and endoscopic therapy for Dieulafoy lesions. Although surgical intervention may be required after failed endoscopic therapy, endoscopy is still an important adjunct for management because a nonbleeding Dieulafoy lesion may be undetectable through a gastrotomy. Because of this potential problem, a combined endoscopic and surgical approach has been adopted. The vascular malformation can be marked with India ink through the endoscope. Rebleeding after endoscopic therapy occurs in 11-15% of cases, with most cases of rebleeding controlled at repeat endoscopy (Reilly, 1991). The 30-day mortality rate from the study by Norton et al was 42%, which is a reflection of the severe comorbid conditions associated with patients who have bleeding from a Dieulafoy lesion. Angiodysplasia Angiodysplasia of the upper GI tract accounts for 2-4% of bleeding lesions (Stabile, 2000). The condition is also a cause of lower GI bleeding in 6% of cases (Corson, 2000). The lesion is a vascular malformation that represents an abnormal dilation of mucosal and submucosal vessels. Histologically, angiodysplasias are dilated, thin-walled vascular channels that appear macroscopically as a cluster of cherry spots. When located in the upper GI tract, they most commonly involve the stomach and duodenum. The lesions can be acquired or congenital, as in hereditary hemorrhagic telangiectasia and Rendu-Osler-Weber syndrome. The acquired lesions are commonly found in patients with chronic renal failure requiring hemodialysis and with aortic valvular disease (especially aortic stenosis). Other diseases, such as cirrhosis and von Willebrand disease, are associated with a higher frequency of angiodysplasia.

29 Most lesions are smaller than 1 cm in diameter and can be multiple in 66% of patients (Stabile, 2000). Bleeding from these lesions can range from occult blood loss to lifethreatening hemorrhage. Because the lesions are small and superficial, endoscopic therapy is highly successful. Endoscopic methods and devices used for hemostasis include lasers, contact heat probes, electrocoagulation, and injection therapy. The contact probe coagulators have been the most common form of endoscopic treatment because of their proven success and ability to target a bleeding lesion tangentially. Recurrent bleeding can occur from the mucosal injury caused by the coagulation. To overcome the possibility of a delayed hemorrhage, endoscopic band ligation has been applied for hemostasis in nonvariceal GI bleeding, including angiodysplasias (Jo, 1999). A recent prospective study comparing endoscopic band ligation to electrocoagulation for nonvariceal UGIB revealed that endoscopic band ligation is safer and faster and provides better results in achieving hemostasis (Matsui, 2002). When endoscopic techniques fail, surgical resection becomes necessary. When pangastric involvement is the source of bleeding, a total gastrectomy may be required. Available nonsurgical options include angiography with catheterdirected vasopressin. Combined hormonal therapy with estrogen and progesterone is reportedly beneficial for patients in whom the diagnosis is unknown and vascular lesions are suggested. Aortoenteric fistula An aortoenteric fistula results from the erosion of the aortic graft into the bowel lumen, usually the third or fourth portion of the duodenum. The result is a direct communication between the aortic graft lumen and the bowel lumen. Most aortoenteric fistulas involve the proximal aortic anastomotic suture line. Initially, patients most often present with a self-limiting sentinel hemorrhage that is then followed by an exsanguinating massive GI bleed. For the warning lesser sentinel bleed in a patient with a history of an abdominal aortic aneurysm repair or a known aortic aneurysm, the diagnosis of a graft-enteric fistula should be considered. An upper endoscopy is the procedure of choice to help diagnose the fistula. It should be performed to the ligament of Treitz. Upper endoscopy findings also help exclude other sources of UGIB. CT scanning is useful because images may reveal thickened bowel, perigraft fluid collection, extraluminal gas, or inflammatory changes in the area of the duodenum and aortic graft. Angiography requires active bleeding (1 mL/min) to be diagnostic. Once the diagnosis is confirmed or seriously considered, emergency surgical intervention is required. In most instances, the aortic graft is removed after debridement and closure of the duodenum, followed by an extra-anatomic vascular bypass to bypass the ligated aorta and revascularize the lower extremities. The perioperative mortality rate is 25-90%, and major complications are common. Published opinions state that graft excision is not necessary as long as no gross contamination and purulence are present at the time of laparotomy (Young, 1999). Under these circumstances, antibiotics are administered longterm. Another option emerging in the surgical literature is the use of endovascular stents to repair the fistula (Deshpande, 1999). Endovascular stent management is

30 technically feasible and may be used as a bridge to more definitive treatment after hemodynamic stabilization in high-risk surgical patients. Stent grafting immediately controls hemorrhage; however, the device is placed in an infected field. As a result, adjunctive measures, such as long-term antibiotics, percutaneous drainage, and bowel diversion, may be required (Burks, 2001). Although endovascular stents have been shown as effective therapy for AEF, case reports are now emerging describing AEFs in AAA patients treated initially with stent grafts as well (Abou-Zamzam, 2003). Conclusion Historically, UGIB has been a surgical problem. Today, patients are initially and nonoperatively treated with endoscopic and angiographic methods to help diagnose and control the bleeding source. Because of these advances, surgical intervention is now more targeted to the specific diagnosis and can be preoperatively planned. Although these technological advances have decreased the need for surgical intervention, a need for the traditional operative procedures will always exist, especially in life-threatening instances.