| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Vascular and Interventional Radiology |
1 From the Departments of Radiology and Medical Imaging (L.D., P.V., M.K.), Gastroenterology (M.D.V.), Abdominal Surgery (P.P., R.T.), Medical Informatics and Statistics (G.V.M.), and Intensive Care (J.D.), University Hospital of Gent, De Pintelaan 185, B-9000 Gent, Belgium. Received November 24, 1999; revision requested January 14, 2000; revision received June 9; accepted July 11. Address correspondence to L.D. (e-mail: Luc.Defreyne@rug.ac.be).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Results of 40 embolizations in 91 patients who underwent arteriography for acute nonvariceal GIH were retrospectively studied. GIH was upper, lower, or transpapillar (hemobilia, pancreatic duct bleeding). Clinical parameters and embolization data were assessed for clinical success and in-hospital survival.
RESULTS: Technical success (bleeding target devascularization) was achieved in all patients except one with upper GIH (39 [98%] of 40 patients). No bowel complications occurred. One partial liver lobe and one partial spleen infarction were noted. Five (13%) of 39 patients bled again within 3 days; all had upper GIH (P = .049). Clinical success (no rebleeding after 30 days) was achieved in 32 (82%) of 39 patients. Clinical success occurred in 13 (68%) of 19 patients with upper GIH, in 10 (91%) of 11 with lower GIH, and in all with transpapillar GIH (P = .084). Mortality rate was 28% (11 of 40 patients), equally spread over upper, lower, and transpapillar GIH (P = .87). Blood loss (hemoglobin level < 80 g/L, P = .041), use of packed cells (P = .049) and fresh frozen plasma (P = .006); shock (P = .047); and corticosteroid use (P = .036) were related to rebleeding. Shock (P = .039) and use of fresh frozen plasma (P = .003) before embolization and rebleeding (P =.012), coagulopathy (P = .007), and need for surgery (P = .03) after embolization were strongly correlated with mortality.
CONCLUSION: Embolization is an effective first approach with lower and transpapillar GIH after endoscopy; it was less effective with upper GIH.
Index terms: Endoscopy, 70.1269 • Gastrointestinal tract, angiography, 70.124 • Gastrointestinal tract, hemorrhage, 70.719 • Gastrointestinal tract, interventional procedures, 70.1264 • Gastrointestinal tract, surgery, 70.45
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The outcome of surgery in acute GIH is determined by the knowledge of the bleeding point (7), the type of operation (8), and the general condition of the patient (9). If the bleeding source is unclear or if the patient is a poor surgical candidate, the use of visceral arteriography has been advocated to identify intraluminal leakage of contrast medium or vascular disease and, at the same time, to provide control of the GIH with embolization or selective infusion of vasopressin (7,10). Recently in Europe, radiologists seem to have become more reluctant to administer local vasoconstrictive therapy because of a high percentage of rebleeding, the pharmacologic side-effects, and the drawbacks of using an indwelling catheter for hours or even days (11,12).
As early as 1978, Bookstein et al (13) suggested that, in the management of lower GIH, embolization might become an acceptable alternative to laparotomy. A number of articles (14,15) on the embolization of acute arterial bowel bleeding have reported technical success rates of 80%–95% and 30-day failure rates of 0%–42%. In most surveys, the number of patients is small (more than half of these studies contained occasional observations, only a dozen studies had more than 10 cases). Moreover, the outcome is often biased by the recruitment of patients on the basis of surgical risk, which is rarely adequately specified (16), or by the strategy of deliberate postembolization resection. Although embolotherapy of transpapillar bleeding seems less controversial, similarly diverse results are present in the articles on arteriobiliary or arteriopancreatic fistulas (17).
With these observations in mind, we have been using embolization in our institution as the first approach for endoscopically unmanageable acute nonvariceal GIH, irrespective of its origin or the patient’s risk status. The purpose of the present retrospective study was to determine the technical and clinical results that we achieved in 4 years.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Patient Selection
With a cohort of 91 patients who underwent arteriography for acute nonvariceal GIH unmanageable with endoscopy, we retrospectively analyzed the findings of all patients who had been treated with embolization. These patients had been referred for arteriography (not for immediate surgical exploration) on the advice of the referring endoscopist (M.D.V.) and the surgeon (P.P.) on duty. Embolization was then performed if arteriography revealed a bleeding source. In the case of a negative gastroduodenal arteriographic finding combined with an endoscopic finding of diffuse bleeding in either the stomach or the duodenum, a prophylactic, or blind, embolization of the left gastric artery or gastroduodenal artery was an option (18,19).
Preembolization endoscopy was omitted in three patients because of severe neutropenia (patient 26), bleeding immediately after a percutaneous bile duct procedure (patient 38), and emergency reasons (patient 15).
Clinical Data
All clinical, laboratory, and endoscopic data, as well as the care given and outcome data, were obtained from the medical records. Coagulopathy was defined as thrombocytopenia with a platelet count less than 50,000 per cubic millimeter (5 x 1010 per liter) and/or a prothrombin time (expressed in percentages) less than 50% of the coagulation activity of a normal reference plasma. Intensive care therapy for reasons other than GIH was assumed to be a supplementary risk factor.
Acute GIH was defined as hematemesis, melena, or red per anal bleeding within 24 hours prior to embolization. Blood loss was quantified by measuring the hemoglobin concentration and the amount of blood products transfused 0–2 days prior to embolization. The number of transfused units of packed cells, fresh frozen plasma, and platelets before and after embolization was recorded. The severity of hypovolemia and shock was graded as a function of intravenous administration of plasma-expanding solutions and catecholamines. Cardiovascular shock parameters at the time of arteriography were less adequate for the evaluation of hypovolemia because most of the patients were more or less hemodynamically stabilized during transport.
Endoscopic findings were classified as normal, active bleeding with lesion, active bleeding without visible lesion, old blood and clot, or transpapillar bleeding. The bleeding risk of ulcers was estimated by applying the Forrest classification (Ia–b, spurting or oozing artery; IIa–c, visible vessel, adherent clot, black base; III, no stigmata of bleeding) (20).
The cause of GIH was traced by consensus analysis (L.D., M.D.V., P.P.) of the history and all imaging findings and histologic and autopsy findings, if available. Bleeding was defined as upper or lower GIH if the cause was located either proximal or distal to the ligament of Treitz. Transpapillar GIH was defined as either hemobilia or pancreatic-duct bleeding.
Methods and Techniques of Embolization
Celiac and superior mesenteric arteriography was performed transfemorally with a 5-F Cobra- or Simmons-type catheter and with 30–40 mL of a nonionic contrast medium (Iopamidol; Nycomed Amersham, Oslo, Norway) injected at a flow rate of 5–7 mL/sec. Inferior mesenteric arteriography required lower doses of contrast medium (20 mL) and injection rates (2.5 mL/sec). Images were obtained with the digital subtraction technique. If necessary, 20–40 mg of butylhyoscine (Buscopan; Boehringer Ingelheim, Brussels, Belgium) was infused as a bolus to prevent bowel motion artifacts. Pharmacoarteriography with the use of anticoagulants, vasodilators, or fibrinolytic agents to provoke contrast medium extravasation was not performed.
The arteriographic diagnosis was extracted from the original reports made by two authors (L.D. and M.K.). Active GIH was proved and localized unequivocally by demonstrating extravasation of contrast agent. Vascular anomalies, such as pseudoaneurysm, angiodysplasia, or arteriovenous malformation, were considered to be definitive bleeding sources. Less specific but suggestive findings were larger mucosal blushes with abnormal vessels (indicating tumor) and prolonged mucosal contrast spots (indicating mucosal focus of inflammation or small vascular malformations).
The choice of embolic agent was case-dependent and was ultimately left to the discretion of the radiologists (L.D., M.K.). In general, three major principles of embolization were respected. First, in the event of contrast medium extravasation, the mural artery (gastroduodenal tract) or the vas rectum (small and large gut) was catheterized as close as possible to the rupture site and then occluded. If direct entry into the bleeding artery was not possible, it was occluded by means of flow-directed injection of the embolic agent.
Second, in distinct cases of acute upper GIH with negative arteriographic findings, blind embolization of the left gastric artery trunk and/or gastroduodenal artery was performed. This was performed preferably with the use of resorbable 2–4-mm gelatin sponge plugs. If distal catheterization of the gastroduodenal artery failed, then a catheterization of the inferior pancreaticoduodenal arteries via the superior mesenteric artery was performed to synchronously occlude the lower arcade.
Third, with hemobilia or pancreatic-duct bleeding, the ruptured vessel segment (pseudoaneurysm) was trapped either by means of proximal and distal coil placement with the sandwich technique or by means of flow-directed (or, rarely, endosaccular) injection of a highly concentrated tissue adhesive.
Thirty minutes after embolization, follow-up arteriography was performed to confirm technical success. The patients then received supportive medical therapy in the intensive care unit. They were observed for ischemic complications in particular. All patients who underwent gastroduodenal or colonic embolization also underwent follow-up endoscopy 1–3 days later. Patients with subsequent bleeding were referred for surgery or endoscopy but not for repeat arteriography.
Informed consent for embolization was obtained from the conscious patient as far as the emergency permitted. Otherwise, the immediate family was informed. This retrospective study was approved by the Ethics Committee of our hospital (project 99/150).
Data Analysis
Embolization success rates were analyzed according to the guidelines recently published by the Society of Cardiovascular and Interventional Radiology (21). Two radiologists (L.D., M.K.) assessed in consensus the result of embolization immediately after the intervention (technical success). Four investigators (L.D., M.D.V., P.P., J.D.) evaluated in consensus the 30-day outcome (clinical success) and the follow-up findings until death or until the end of the study observation time (complete clinical or long-term success). We included an extra, early 0–3-day observation period to detect therapy-related failures (ie, recanalization of the embolized artery, refilling through collateral vessels).
The embolization was defined as technically successful if follow-up arteriography revealed devascularization of the target vascular lesion or disappearance of contrast material extravasation. Blind embolization was technically successful if the embolized artery did not opacify on the follow-up angiogram. Rebleeding was assessed by checking clinical parameters (clearing of aspirates from the nasogastric tubes, no passage of blood in the stool) and hematologic parameters (hemoglobin concentration, packed cell transfusions).
Technical arteriographic and embolization data were evaluated for early 3-day rebleeding and in the 30-day clinical-success study (L.D., G.V.M.). Angiographic localizations were first analyzed as a function of their anatomic sites (eg, stomach, duodenum, jejunum) and further as subgroups (upper or lower gastrointestinal tract GIH, transpapillar GIH, no visible disease). The effect of active bleeding at arteriography on the outcome was evaluated by comparing results in patients with contrast material extravasation and/or pseudoaneurysm with those of patients presenting with structural abnormalities or without abnormal findings.
Disappearance of contrast medium extravasation during superselective catheterization was carefully documented. Embolized arteries were first stratified as a function of their anatomy and were then assigned to one of three subgroups (upper or lower gastrointestinal tract arteries, arteries of the liver and/or pancreas). The catheterization technique was recorded as selective (with a 5-F catheter), superselective (with a microcatheter system), or a combination of both. Embolic agents were studied for their influence on outcome both separately (polyvinyl alcohol particles, absorbable gelatin sponge, N-butyl-cyanoacrylate [glue], Ethibloc [Ethicon, Hamburg, Germany], stainless steel coils, platinum microcoils) and in combination if more than one agent was used in a particular patient.
Because survival may depend on factors other than the successful arrest of GIH, embolization success rates and patient survival were analyzed separately. We used in-hospital survival rates to assess postprocedural morbidity and mortality.
The Pearson 
2 test or Fisher exact test was used to compute the effect of categorical demographic, arteriographic, and interventional data on therapy outcome and survival. The numeric data were analyzed with the nonparametric Kruskal-Wallis analysis of variance or the Mann-Whitney U test, as was appropriate. Analyses of relevant subgroups (ie, upper, lower, and transpapillar GIH) were performed with the same statistical procedures.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
Embolization Techniques Used
Arteries of the upper alimentary tract were occluded by using a variety of embolic agents with different techniques (Table 2). In seven (35%) of 20 patients with upper GIH, a combination of two or more embolic agents was used.
Lower GIH (n = 11) was treated with superselective catheterization of the vas rectum and injection of polyvinyl alcohol particles in all patients except two (Table 2). In one patient with ischemic cecal bleeding, a 2 x 5-mm platinum microcoil was placed proximally in a cecal branch because arteriography still revealed focal mucosal staining after polyvinyl alcohol particle injection. The other patient had lower GIH from a rectal pseudoaneurysm induced by laser therapy of a tubular adenoma. This pseudoaneurysm was embolized with glue.
Bleeding from a pancreatic pseudoaneurysm was encountered twice (Table 2). One huge gastroduodenal aneurysm that expanded into the pancreatic head was embolized with platinum microcoils by using a sandwich technique. One splenic arterial aneurysm in the pancreatic tail was endosaccularly embolized with glue. Of seven patients with hemobilia (Table 2), two were treated with stainless steel coils pushed through a 5-F catheter; in one case, a sandwich technique occlusion was performed, and in the other, distal artery occlusion combined with proximal injection of absorbable gelatin sponge. In the remaining five patients with an arteriobiliary fistula, the vascular liver anatomy required superselective microcatheterization.
Technical Success
The technical success rate of embolization was high (39 [98%] of 40 patients). The only failure occurred in a patient (patient 19) with postsphincterotomy bleeding. Celiac and mesenteric arteriography revealed contrast medium extravasation from a superior branch of the anterior pancreaticoduodenal artery. Only the gastroduodenal approach was feasible, but it was insufficient because contrast medium extravasation persisted at mesenteric arteriography. The patient underwent surgical hemostasis successfully.
In six (29%) of 21 patients with contrast medium extravasation on main-trunk arteriographic images, leakage was no longer visible after superselective catheterization. The same was observed in a patient with angiodysplasia and in one with an unclear prolonged mucosal contrast agent spot (later clarified as an arteriovenous malformation). In these patients, correct microcatheter positioning was possible by comparing the images of a superselective series with images of the main trunk series. This comparison allowed distal and/or flow-directed occlusion, even without contrast medium leakage being viewed. In another patient (patient 10), duodenal bleeding became evident only after contrast medium injection in a selective wedged position.
Techniques used with lower GIH differed from those used with upper and transpapillar GIH because of the consistent use of a microcatheter to achieve a distal superselective position (P = .042) and because of the injection of polyvinyl alcohol as the occlusive agent (P = .001).
Periprocedural Complications
Technical complications occurred in two patients. In patient 16, the branch bridging the upper and lower pancreaticoduodenal arcades ruptured during contrast medium injection after delivery of the first stainless steel coil. Continuation of the procedure with placement of a second coil by using the sandwich technique solved the problem. In patient 39, the last coil was stuck in a kinked microcatheter until the end of the procedure. Manipulation ruptured the microcatheter, which was left behind, uneventfully, in the hepatic artery.
Postembolization bowel ischemia was not observed clinically or at endoscopy. Occasionally, the endoscopist reported nonspecific erythema of the gastroduodenal or colonic mucosa after embolization.
Visceral organ infarction complicated recovery in three patients. Partial infarction of the left liver lobe ensued from embolization of the left hepatic artery for an iatrogenic arteriobiliary fistula in a patient with cholangiocarcinoma and left portal vein invasion (patient 38). Conservative treatment was successful. In a patient (patient 40) with a distal pancreatic pseudoaneurysm, embolization with glue provoked partial splenic infarction, which was treated with splenectomy. Recovery was complicated by a contaminated pseudocyst. One patient (patient 29) survived a cardiac infarction. In another patient (patient 11), reflux of a droplet of glue into the right hepatic artery occurred during gastroduodenal embolization but was clinically uneventful.
Early 3-day Failures
Twenty-nine (74%) of 39 patients who underwent technically successful embolization had no clinical evidence of and hence received no treatment for GIH after 3 days (Table 2). Ten patients received either early surgical (n = 4) or endoscopic treatment (n = 6), although clinical evidence of recurrent GIH was not always conclusive. Early surgery was performed in one patient with overt duodenal bleeding and in three with endoscopically unmanageable upper GIH recurrences. In one other patient with rebleeding, an oozing bulbar ulcer (Forrest Ib) was revealed at endoscopy. He was treated successfully with sclerotherapy. Five other patients underwent endoscopic therapy for nonbleeding lesions within 3 days after embolization (Table 2).
The five clinical cases of rebleeding were confirmed by the use of significantly more units of packed cells transfused (mean, 5.6 vs 1.2 units; P = .026) and by the lower hemoglobin concentration (mean, 81 vs 93 g/L; P = .001) after embolization. Therefore, the early failure rate is 13% (five of 39 patients). Upper GIH recurred significantly more frequently (five [26%] of 19) within the first 3 days than did lower (0 of 11) or transpapillar (0 of 9) GIH (P = .049).
Hemoglobin concentration below 80 g/L (P = .062), shock (P = .058), and corticosteroid use (P = .07) prior to embolization were more frequently observed in patients with early failure than in other patients.
Clinical Success Rate
We observed two instances of rebleeding between days 4 and 30. One patient had melena from a bleeding gastric ulcer on day 1; he was treated with endoscopy. The other patient had from Henoch-Schönlein disease with diffuse GIH on day 8 that was resistant to treatment with a combined enteroscopic and surgical approach. Adding the early 3-day failures to these 4–30-day failures, we obtained an overall 30-day recurrence rate of seven (18%) of 39, or a clinical success rate of 82%.
None of the arteriographic or interventional parameters had a substantial influence on either technical or clinical success. At 30 days, rebleeding rates had increased to six (32%) of 19 patients (including the five with early instances of rebleeding) with upper GIH and to one (9%) of 11 patients with lower GIH, with no increase (0 of 9 patients) in transpapillar GIH. This difference in rebleeding rates obviously remained, but it was only on the borderline of significance (P = .084). Compared with lower and transpapillar GIH, upper GIH was manifested with a lower hemoglobin concentration (63.2% with <80 g/L, P = .007) and with higher amounts of packed cells transfused (mean, six units per 48 hours vs three and two units per 48 hours; P = .024).
In the subset of patients with upper GIH, early failure occurred more frequently with duodenal bleeding (four [40%] of 10 patients) than with gastric bleeding (one [11%] of nine patients), but the difference was not statistically significant (P = .3). After 30 days of observation, only one additional instance of gastric rebleeding (two [22%] of nine patients) was noted. As for the arteriographic findings, only contrast medium extravasation was more frequently demonstrated with duodenal bleeding (eight [80%] of 10 patients) than with gastric bleeding (three [33%] of nine patients) (P = .07). However, the presence of contrast medium extravasation at arteriography did not correlate with recurrent GIH in either the whole study group (P = .29) or the subgroup with upper GIH (P >.99).
Disappearance of contrast medium extravasation with superselective catheterization did not influence early failure (P = .2) or clinical success rates (P = .61).
Significant initial clinical risk factors for overall 30-day failure were shock (P = .029) and corticosteroid use (P = .037) but not coagulopathy (P = .65). Patients experiencing clinical failures had received significantly higher amounts of packed cells (9.5 units ± 6.5 [mean ± SD] vs 5.5 units ± 6.8; P = .049) and fresh frozen plasma (5.3 units ± 4.2 vs 1.1 unit ± 1.8, P = .006). However, these patients showed no difference from those with clinical success in terms of hemoglobin concentration (78 g/L ± 26 vs 88 g/L ± 24, P = .3).
Complete Clinical Success Rate
Long-term recurrent GIH with surgical intervention occurred in one patient each at days 40 and 132. In patient 26, ischemic colitis was demonstrated at histopathologic examination. In patient 22, repeat arteriography uncovered a previously embolized nonspecific jejunal contrast material spot as a small arteriovenous malformation, which was resected.
In-Hospital Survival
The outcome of surgery for early rebleeding was unfavorable. Within 3 days, one surgical patient died; between 4–30 days, three other surgical patients died (persisting GIH, n = 3; sepsis, n = 1). In one, concomitant portal hypertension and cirrhosis was uncovered at arteriography. One patients who successfully underwent embolization (Henoch-Schönlein vasculitis) bled again on day 8 and did not survive.
In the group with clinical success, six patients died from causes unrelated to bleeding: hepatic failure (n = 1), cardiac failure (n = 2), acute respiratory distress syndrome (n = 2), and septic shock (n = 1).
The overall in-hospital mortality was 28% (11 of 39 patients). Four (36%) of 11 deaths were directly related to persisting or recurrent GIH despite all interventions, and three (75%) of four instances of GIH were located in the upper gastrointestinal tract. Outcome by location revealed six deaths among patients with upper GIH (32%), three deaths among patients with lower GIH (27%), and two deaths among patients with transpapillar GIH (22%). The mortality rates in upper, lower, and transpapillar GIH were similar (P = .87).
Survival was inversely correlated with preembolization clinical parameters, such as shock therapy (P = .039) and transfused fresh frozen plasma (P = .003), and with postembolization clinical factors, such as surgery for endovascular failure (P = .004), coagulopathy (P = .007), and shock therapy (P = .002). Early 3-day and clinical 30-day failures were associated with a higher mortality (P = .017 and P = .012, respectively). The variables reflecting rebleeding and coagulopathy after embolization—decrease in hemoglobin concentration (P = .014) and units of packed cells (P = .002), fresh frozen plasma (P = .015), and platelets (P = .008) transfused after embolization—were directly correlated with in-hospital mortality.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Technical Success Rate
Recently developed microcatheter and guidewire systems permit a close endovascular approach to most GIHs. Distal catheterization is limited by vessel tortuosity, bowel peristalsis, and, especially, vessel-wall reactivity (22). Vasospasms preclude entrance into the bleeding artery and may provoke bleeding arrest (23). By repositioning the microcatheter and waiting patiently, normal flow will usually reconstitute. Moreover, contrast material injection from a more proximal position may disclose the bleeding site on or through an adjacent artery. Our experience with local vasodilators was not extensive, and there results were not conclusive. Intermittent contrast material extravasation may also reflect spontaneous bleeding arrest (24). Whenever contrast material leakage disappears, main trunk arteriographs should be meticulously studied again to determine the point for safe embolization. Because extravasation was initially localized, we found no indication for the use of anticoagulants or fibrinolytics, as other authors have suggested (25,26).
Complication Rate
The absence of bowel ischemia indicates that we delivered the embolic material in the right amount and at the right place. Although we narrowed the risk area by performing embolization into the ruptured mural artery, there is experimental evidence to indicate that delivery of polyvinyl alcohol particles of 100 µm or larger from a more proximal site is effective as well and that this delivery does not disturb blood distribution to the adjacent bowel (27). Recently, several authors (22,23,28) have confirmed that polyvinyl alcohol particles and platinum microcoils are safe and curative in lower GIH embolization. Polyvinyl alcohol particles, injected just proximally to the vas rectum, seem to flow along the path of least resistance, particularly the ruptured vas rectum, and reduce the risk of bowel ischemia. Nicholson et al (29) reported difficulties in negotiating the vas rectum with subsequent coil placement in the marginal artery. In this series, proximal occlusion was considered responsible for ischemic colonic strictures in three (21%) of 14 patients; all were treated conservatively. Embolization in the vas rectum does not seem to increase ischemic complications contrary to what was previously thought (30).
Special skill is required to assess adequate dilution of liquid embolic agents to control vascular penetration and preserve bowel perfusion. It is no wonder that there are few reports (31,32) of using glue for GIH embolization. Although we successfully treated two duodenal and one rectal GIHs with glue, we remain reluctant to this technique because of the higher risk of bowel infarction and stenotic complications (31).
Gas-building tissue necrosis caused by visceral organ infarction (patients 38 and 40) was difficult to differentiate from an abscess, which hampered therapeutical decision making. In both patients with necrosis, the risk of embolization outweighed the risks of emergency surgery.
Early Rebleeding
We found no statistical evidence that the method of embolization and the embolic agents used affected the early outcome. However, all five (13%) early bleeding recurrences were located in the gastroduodenal region. The upper gastrointestinal tract emerged as a high-risk region for early failure (P = .049). Early recanalization of the gastric or duodenal arteries after embolization with use of absorbable gelatin sponge could be responsible for recurrence. However, only one (20%) of five patients bled again after occlusion with absorbable gelatin sponge, whereas failure rate of 30% (three of 10 patients) was observed with permanent occlusive polyvinyl alcohol (Table 2). One may argue that a normal arteriogram (cases with absorbable gelatin sponge embolization) allows prediction of a favorable outcome because bleeding had already ceased. On the other hand, contrast material extravasation could be an additional risk factor for clinical failure. We found no correlation between extravasation on arteriograms and rebleeding.
Refilling through the extensive collaterals in the duodenal blood supply may explain early rebleeding after technical success. We suggest that inflammatory or ischemic reactions after embolization trigger vasodilation of the intramural collaterals. Refilling could be effectively eliminated by occluding the gastroduodenal artery with a penetrating liquid embolic agent, such as glue, or by using a superselective approach to occlude the bleeding intramural artery (31). However, the use of glue is limited by the risk of ischemic complications. Alternatively, a superselective approach implies the presence of extravasation and a favorable vascular anatomy. Although synchronous embolization of the pancreaticoduodenal arcades is theoretically obvious, we, as well as other authors (33), did not obtain the expected results with it.
Obviously, clinical factors play an important role in early failure of embolization in upper GIH. First, embolization, like surgery, does not treat the peptic ulcer disease. Gastric acid suppression is required for ulcer treatment. To this point, Kolkman and Meuwissen (34) recommended restricting surgery to undersewing the bleeding ulcer artery, circumventing collateral refilling, as is the intention of embolization. In their opinion, pyloroplasty and vagotomy should become obsolete because the key factor in peptic ulcer disease is not gastric acid production but the pathogenicity of Helicobacter pylori. The combination of gastric acid suppression with medical eradication of H pylori has been effective in the prevention of rebleeding (35).
Second, upper GIH is more profuse and resistant to hemostasis than is lower GIH. In our series, low hemoglobin concentration (<80 g/L; P = .062) and shock (P = .058) were more frequently observed in patients with early failure, which points to the severe and brisk character of upper GIH. Signs of shock and active bleeding at admission are known risk factors for rebleeding after endoscopic therapy (36–38); hence, they are probably risk factors for early recurrence after embolization as well.
Nonetheless, we are disappointed by our high early failure rate in duodenal ulcer GIH, when compared with the 0%–20% rate reported elsewhere (14,15,32,33). In our opinion, these results in the literature are somewhat optimistic. Recently proposed quality improvement guidelines (21) for embolization recommend a far lower clinical success threshold of 60% for bleeding duodenal ulcer. This threshold reflects the results of the largest series of duodenal embolizations yet published (31).
In lower GIH, polyvinyl alcohol embolization resulted in complete early success, thereby challenging the promising results recently obtained with other embolic agents (22,23,29). From this experience and on the basis of the statistical evidence, we consider polyvinyl alcohol particles (150–250 µm) the agent of choice for distal flow-directed occlusion in lower GIH. The only other report (28) of polyvinyl alcohol embolization for lower GIH had a recurrence rate of 30%. Although the exact location of the recurrent bleeding is arguable, we suggest that the larger size (
255–350 µm) of the polyvinyl alcohol particles injected in that study could have blocked blood flow more proximally in the marginal artery. In such cases, collateral circulation between vasa recta will persist and allow refilling of the bleeding artery. Peck et al (23) proposed a similar mechanism for rebleeding after embolization with absorbable gelatin sponge and microcoils. In a series of coil embolizations for colonic bleeding, Nicholson et al (29) reported two (14%) of 14 early recurrent bleeding that required surgery, one with Crohn colitis and one from a hemangioma. From this article, it is not clear whether the site of the coil delivery allowed collateralization or if the rebleeding was due to the nature of the bleeding source.
Clinical Success Rate
In patients without early rebleeding, cardiovascular hemodynamics are reestablished, promoting endogenous repair of the rupture site by means of thrombus formation and organization. Therefore, rebleeding after 3 days is more likely to be related to clinical factors than to the embolization technique. Indeed, in our series, both 4–30-day rebleeding occurred because of malignancy and vasculitis.
Patients receiving vigorous hypovolemic shock therapy and high quantities of packed cell transfusions (mean, 9.5 units per 48 hours) before embolization had a significantly higher risk for 30-day rebleeding (39). In this group of patients (which includes those with 3-day early failures), resuscitation seemed insufficient to reestablish hemostasis despite technically successful embolization and rescue surgery.
Corticosteroid intake was considered to be a statistically significant risk factor for rebleeding after embolization probably because of its retarding wound-healing effect. Corticosteroid use is more often encountered in inpatients with bleeding than in those with primary-referred GIH (38,40), but it has not yet been reported to be an independent risk factor for rebleeding.
All four patients receiving postembolization prophylactic sclerotherapy of Forrest class II lesions remained stable during follow-up. The combination of embolization with subsequent endoscopic control and adjuvant sclerotherapy seems promising for the control of upper GIH (41).
In 15%–20% of patients with obscure lower GIH treated conservatively, short- and long-term rebleeding from uncovered disease can be expected (42). We observed a similar rebleeding rate of 27% in three of 11 patients who needed surgery, two emergently and one semiurgently. It remains remarkable that a diagnosis was never established in three of 11 patients with lower GIH despite all of the imaging studies. Although one of these patients died from concomitant disease, two survived without recurrence.
In the past, surgery for acute hemobilia or pancreatic duct bleeding had high morbidity and mortality rates (17,43). Nowadays, embolization has gained widespread acceptance as the procedure of choice in transpapillar bleeding (43–45). Contrary to other reports (44,46), clinical success in this series was complete, which suggests that our approach seems to suffice for arresting bleeding.
In-Hospital Survival
Overall in-hospital mortality was 28% (11 of 39 patients) with similar death rates with upper (32%), lower (27%), and transpapillar (22%) GIH. Rebleeding in upper GIH was highly correlated with mortality (36). All four patients with rebleeding died after rescue surgery, which failed to stop GIH in three of the four. The mortality of secondary surgery exceeded even the 50% death rate with emergency surgery (47,48). All patients who died after surgery had developed severe coagulopathy postembolization. Hypocoagulability was significantly correlated with mortality and was provoked by the excessive blood loss before surgery was started (49). In other environments, patients with such a severe shock and coagulopathy would not have been considered candidates for surgery.
In recurrent upper GIH, should repeat arteriography and repeat embolization be advised (50)? Certainly, findings at repeat arteriography may clarify a previously nonspecific bleeding source. Also, findings at repeat arteriography may clarify the cause of the early failure of embolization and provide justification for repeat intervention. However, in view of our inconsistent results with embolization of upper GIH, we are still reluctant to recommend this as the strategy of choice. Early repeat arteriography still has a substantial risk that no bleeding source will be found and therefore may further delay hemostasis.
In lower and transpapillar GIH, mortality was caused by rebleeding in one and by comorbidity in all five fatal outcomes. These patients had a poor risk profile and were not surgical candidates.
In endoscopically unmanageable lower and transpapillar GIH, we advise embolization as the first approach because it offers definitive and safe nonsurgical healing in a large number of cases. Because of compromising clinical factors and vascular complexity, we found embolization of upper GIH to be less effective. Therefore, further studies are required before embolization can be proposed as the routine first approach for endoscopically unmanageable nonvariceal upper GIH. The overall mortality rate remained high; it was directly related to the rebleeding rate and the severity of loss of blood but not to the bleeding site.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, L.D.; study concepts, L.D., M.D.V., P.P.; study design, L.D.; definition of intellectual content, L.D., M.D.V., P.P., M.K.; literature research, L.D., P.V., J.D., R.T.; clinical studies, L.D., P.V.; data acquisition, L.D., P.V., J.D., R.T.; data analysis, G.V.M., L.D.; statistical analysis, G.V.M.; manuscript preparation, L.D.; manuscript editing, L.D.; manuscript review, P.V., M.D.V., J.D., P.P., G.V.M., M.K.; manuscript final version approval, L.D.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  G. A. Poultsides, C. J. Kim, R. Orlando III, G. Peros, M. J. Hallisey, and P. V. Vignati Angiographic Embolization for Gastroduodenal Hemorrhage: Safety, Efficacy, and Predictors of Outcome Arch Surg, May 1, 2008; 143(5): 457 - 461. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. J. Laing, T. Tobias, D. I. Rosenblum, W. L. Banker, L. Tseng, and S. W. Tamarkin Acute Gastrointestinal Bleeding: Emerging Role of Multidetector CT Angiography and Review of Current Imaging Techniques RadioGraphics, July 1, 2007; 27(4): 1055 - 1070. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
