Angiography in poor-risk patients with massive nonvariceal upper gastrointestinal bleeding

Angiography in Poor-Risk Patients with Massive Nonvariceal Upper Gastrointestinal Bleeding Daniel T. Dempsey, MD, Dana R. Burke, roD, Rebecca S. Reilly, Ba, Gordon K. McLean, SaD, Ernest F. Rosato, MD, Philadelphia, Pennsylvania

The purpose of this retrospective study was to determine the diagnostic and therapeutic usefulness of gut angiography in patients with massive upper gastrointestinal bleeding from a nonvariceal source. All patients (n = 64) in this category who underwent a gut angiogram between 1980 and 1986 were studied. Pre-angiogram endoscopy was attempted in all patients and was nondiagnostic in 14 ( 2 2 % ) . Contrast extravasation at angiography was seen in 25 of 64 patients ( 3 9 % ) , and in over half of these patients endoscopy was nondiagnostic (n = 11) or wrong (n = 3 ) . Attempts to control bleeding in this group by selective arterial embolization (n = 14) or intra-arterial vasopressin (n = 11) successfully averted operation in 13 of 25 patients (52%) and was associated with a 50% reduction in mortality (83% versus 3 8 % ) . Selective embolization of vessels thought to be bleeding on clinical grounds without evidence of contrast extravasation (i.e., " b l i n d " embolization) was not helpful in controlling hemorrhage. Urgent gut angiography in patients with massive upper gastrointestinal bleeding of arteriocapillary source is a useful diagnostic and therapeutic maneuver and warrants continued application in this group of poor-risk patients.

cally [4] in selected populations with gastrointestinal bleeding, there are no large series that evaluate this relatively new technique in poor-risk patients with massive arterial hemorrhage from the upper gastrointestinal tract. Because of an aggressive institutional policy for angiography in patients with upper gastrointestinal bleeding, we are able to report such a series. The purpose of this review is to determine the diagnostic and therapeutic value of gut angiography in acute massive nonvariceal upper gastrointestinal bleeding.

PATIENTS AND METHODS All patients (n = 83) who underwent gut angiography for upper gastrointestinal bleeding at the Hospital of the University of Pennsylvania between 1980 and 1986 were reviewed. Nineteen patients were excluded based upon the following criteria: (1) onset of clinical bleeding (grossly bloody nasogastric aspirate or hematemesis) greater than 7 days prior to angiogram (n -- 2), (2) blood transfusion of less than 4 units whole blood or packed red blood cells prior to angiogram (n = 2), (3) variceal bleeding as defined by positive endoscopy or known cirrhosis and variees with no other bleeding site seen on endoscopy or angiography (n = 15). The remaining 64 patients with acute (within 7 days), massive (greater than 4 units of blood transfused), nonvariceal upper gastrointestinal bleeding form the basis for this study. Endoscopy was performed or attempted in all 64 patients prior to arteriography and the results noted. The following variables were recorded for each patient: age, sex, associated diseases, endoscopic diagnosis, transfusion requirement before and after angiography, angiographic findings, angiographic treatment (selective arterial embolization, selective intra-arterial vasopressin he operative mortality rate for massive upper gas- infusion, none), pre-angiogram laboratory data (protrointestinal bleeding (transfusion requirement of at thrombin time, partial thromboplastin time, platelet least 4 units per 24 hours) remains around 20% [1]. In count, creatinine), and in-hospital mortality. Angiopoor-risk populations, it is much higher [2]. Early and graphic diagnosis was compared with endoscopic diagnoaggressive resuscitation, diagnostic work-up, and opera- sis. All patients with angiographic evidence of bleeding tion is the preferred approach. Unfortunately, accurate (i.e., contrast extravasation) had either selective arterial diagnostic endoscopy is sometimes difficult in the patient embolization (n = 14) or selective intra-arterial vasopreswith massive hematemesis. Furthermore, in some pa- sin infusion (n -- 11). These two forms of angiotherapy tients, the risk of surgical intervention appears prohibitive were considered equivalent, and the choice was deterbecause of concomitant problems, such as overwhelming mined by specific patient anatomy, location of bleeding sepsis, persistent hemodynamic instability, and multiple site, and angiographer preference. Generally, those paorgan compromise. Although gut angiography has been tients in whom a stable superselective catheter position shown to be useful both diagnostically [3] and therapeuti- could be obtained in the angiographically demonstrated From the Departmentsof Surgeryand Radiology,Universityof Penn- bleeding vessels were treated with embolization using sylvaniaSchoolof Medicine,Philadelphia,Pennsylvania. Gelfoam pledgets or stainless steel coils, or both. Selective Requests for reprints should be addressedto DanielT. Dempsey, intra-arterial vasopressin therapy was used in those paMD, TempleUniversityHospital,3401 North BroadStreet, Philadel- tients whose vascular anatomy precluded stable supersephia, Pennsylvania19140. Manuscript submittedOctober25, 1988,revisedMarch 17, 1989, lective catheterization of the target vessels. Some exceptions to this general approach did occur because of and acceptedMarch 21, 1989.

T

9.82 THE AMERICAN JOURNAL OF SURGERY VOLUME159 MARCH1990

ANGIOGRAPHY IN UPPER GASTROINTESTINAL BLEEDING

angiographer or surgeon preference. In addition, about half of the patients in whom a bleeding site could not be conclusively demonstrated angiographically had selective embolization of the vessel thought most likely to be bleeding on clinical and endoscopic grounds (n = 20). However, there was a lack of consensus within the departments of surgery and angiography as to the therapeutic potential of this maneuver, so many patients without angiographic evidence of bleeding received no angiotherapy (n -- 19). The need for an operation after angiography to control bleeding (a clinical decision of the attending surgeon) was noted. No patient died of exsanguination. Data were analyzed by Student's t test for unpaired data and chi-square analysis with Yates' correction. RESULTS The age of the patients was 61 4- 16 years (mean 4SD), and men outnumbered women by approximately two to one (42 men, 22 women). Patients received an average of 9 4- 5 units of whole blood or packed red blood cells prior to angiography, which was performed 1.8 41.3 days after the onset of upper gastrointestinal bleeding. Thus, this group was bleeding at the rate of approximately 5.5 4- 4.7 units of blood per day. The mean serum bilirubin and creatinine levels were 3.7 4- 6.3 mg/dL and 1.9 4- 1.9 mg/dL, respectively. The overall mortality was 41% (26 of 64), and the average length of hospitalization after angiography until death or discharge was 23 4- 29 days. Table I defines this population as a high-risk group. Half the patients were over 65 years old, a third had significant cardiac problems, a fifth had pulmonary, liver, or renal insufficiency, and many were septic as defined by positive blood cultures. Most patients had two or more of these significant risk factors. Most patients were bleeding from duodenal (27%) or gastric (22%) ulcers, or gastritis (22%) (Table I). As mentioned above, patients thought to be bleeding from esophagogastric varices were purposely excluded. Patients were grouped according to whether a bleeding site was seen at angiography (contrast extravasation) and whether angiographic control of hemorrhage was attempted. Group 1 (n = 14) consisted of patients in whom extravasation was seen and selective intra-arterial embolization tried. Group 2 (n = 11) also showed angiographic extravasation, but selective intra-arterial vasopressin was infused (0.2 to 0.4 U/min) by an in-dwelling catheter for 24 to 48 hours. Group 3 (n = 20) showed no evidence of extravasation, but had selective arterial embolization based upon the best possible clinical diagnosis and upon the belief of some angiographers or surgeons that this maneuver might be useful in controlling hemorrhage. Group 4 (n = 19) had no extravasation seen and no angiographic therapy attempted. Results of angiographic and endoscopic evaluation are compared in Table II. Endoscopic results were nondiagnostic (n = 22) or wrong (n = 3) in 25 of 64 patients (39%), usually because of poor visibility secondary to massive hemorrhage. Bleeding was visualized by angiography (conclusive extravasation) in 25 of the 64 patients (39%). No bleeding was visualized

TABLE 1

Associated Problems and Diagnosis In 6 4 P a t i e n t s with Massive Upper Gastrointestinal (GI) Bleeding Patients Risk factors Age > 65 yrs CHF/MI/cardiac surgery ~ Ventilator Cirrhosis Renal insufficiency Positive blood cultures Cancer COPD/pneumonia/ARDS t Stroke Status post GI operation Diagnosist Duodenal ulcer Gastric ulcer Gastritis Mallory-Weiss syndrome Postop GI bleeding Other

n

%

32 21 14 14 13 12 12 7 5 4

50 33 22 22 20 19 19 10 8 6

17 14 14 9 4 6

27 22 22 14 6 9

* CHF = congestive heart failure; MI = recent (<30 d) myocardial infarction; cardiac surgery = recent (<30 d) open heart surgery. t COPD = chronic obstructive pulmonary disease; ARDS = adult respiratory distress syndrome. Made retrospectively with all available data,

TABLE

II

Diagnostic Accuracy of Anglography and Endoscopy In Determining Cause of Bleeding Diai~nosls Anglographic

Extravasation

Endoscopy

Positive Positive Negative Negative

Positive Negative Positive Negative

Patients (n = 64) n %

Total

11 14 ~ 28 11 64

17 22 44 17 100

* In three patients, angiography yielded unequivocal extravasation, suggesting a cause different from endoscopy. In 11 patients, endoscopy was nondiagnostic.

angiographically in 61% of the group. In 11 of the 25 patients in whom extravasation was evident, the anglographic impression was consistent with the endoscopic diagnosis. In another 14 of the 25 patients in whom bleeding was visualized, the endoscopic results were nondiagnostic in 11 and wrong in 3. The site of angiographic extravasation and the eventual diagnosis in the 11 patients with nondiagnostic endoscopy are shown in Table III. Diagnosis was confirmed by repeat endoscopy, operation, or autopsy. As mentioned, in three additional patients, angiography revealed conclusive evidence of an erroneous endoscopic diagnosis. One patient thought to be bleeding from gastric varices and another thought to be bleeding from gastritis on endoscopy both had extravasation from the gastroduodenal artery (duodenal ulcer). A third patient with the endoscopic diagnosis of bleeding

THE A M E R I C A N J O U R N A L OF SURGERY

VOLUME 159

M A R C H 1990

283

DEMPSEY ET AL

TABLE IV

TABLE III

Factors Associated with Success in Patients with Contrast ExtravasaUon

Positive Anglogram and NondlagnosUc Endoscopy (n = 11)* Site of Angiographic Extravasation Left gastric artery (7)

Common hepatic artery (1) Gastroduodenal artery (1) Inferior pancreatlcoduodenal artery (1) Superior mesenteric artery (1)

Diagnosis Gastritis (3) Gastric ulcer (2) Mallory-Weiss (1) syndrome Gastrojejunal anastomosis (1) Hemobilia (1) Duodenal ulcer (1) Duodenal ulcer (1) Marginal ulcer (1)

" Values in parentheses indicate number of patients.

Age > 65 yr < 65 yr Pitressin Embolizatlon Blood transfusion (>8U) Pre-angiography (<8U) Platelet count < 120k > 120k Ulcer No ulcer Died Survived

Operation Averted ( n - 13)

Operation Required (n = 12)

Chi-Square ~

5 8 6 7 4 9 5 8 5 8 5 8

8 4 5 7 9 3 9 3 7 5 10 2

1.02 (NS) 0.06 (NS) 3.28 (p < 0.10) 2.06 (NS) 0.35 (NS) 3.53 (p < 0.10)

" With Yates' correction.

esophageal varices had extravasation from the left gastric artery at the level of the incisura (gastric ulcer). Extravasation was more likely to be seen on angiography in those patients over age 70 (59% versus 32%, p <0.10), in patients requiring 6 units of whole blood or packed cells per 24-hour period (54% versus 30%, p <0.10), and in those with a platelet count below 120,000 per mm 3 (56% versus 28%, p <0.05). All 25 patients demonstrating angiographic extravasation had either selective arterial embolization (n = 14) or selective intra-arterial infusion ofvasopressin (n = 11). Arteries embolized in the group of 14 patients were the left gastric (n --- 9), inferior pancreaticoduodenal (n = 4), gastroduodenal (n -- 3), common hepatic (n -- 1), and right gastric (n -- 1). Some patients had embolization of more than one named vessel. In this group of 25 patients with contrast extravasation, operation was undertaken to control bleeding based upon clinical and angiographic evidence of continued hemorrhage. Operation was required in 12 of 25 patients in this group, and this was defined as a failure of angiotherapy. These 12 patients with unsuccessful angiotherapy tended to be bleeding more profusely than the 13 patients in whom embolization or selective vasopressin averted operation (8.9 4- 7.5 versus 5.6 4- 2.0 units/day, p = 0.10). Angiotherapy was less successful in averting operation (Table IV) in patients receiving more than 8 units of blood prior to the study period (p <0.10). Selective intra-arterial infusion of vasopressin and selective intra-arterial embolization appeared equally successful in averting an operation to control bleeding (55% versus 50%), although patients given vasopressin tended to be bleeding less vigorously than those patients who underwent embolization (5.4 42.5 units/day versus 8.4 4- 6.8 units/day). Patients requiring operation after failed angiotherapy had a higher mortality rate than patients in whom operation was avoided (83% versus 38%, p <0.10). In the 39 patients without evidence of extravasation, selective arterial embolization was undertaken in 20. This was based upon the most probable diagnosis considering the clinical, endoscopic, and angiographic (e.g., gastric 284

THE AMERICAN JOURNAL OFSURGERY

hyperemia suggesting gastritis) findings. This so-called blind embolization usually involved the left gastric artery (n = 15). Other arteries embolized blindly were the gastroduodenal (n = 3), inferior pancreaticoduodenal (n = 2), common hepatic (n -- 1), and right gastric (n -- 1). In 19 of the 39 patients without angiographic extravasation, blind embolization was not performed, usually because of physician preference. After blind embolization, subsequent operation was required in 6 of the 20 patients to control bleeding. In the comparable group without embolization (n -- 19), operation was required in the same percentage of patients (6 of 19). Data summarizing the four diagnostic/therapeutic angiographic groups are presented in Table V, according to whether subsequent operation was required for control of hemorrhage. No patient in the series had gastric necrosis at operation or autopsy. COMMENTS Our findings suggest that in a poor-risk population with massive nonvariceal upper gastrointestinal bleeding, angiography is sufficiently useful diagnostically and therapeutically to warrant continued application in this group. We found evidence of angiographic extravasation in 40% (25 of 64) of our patients. In over half of these patients in whom angiographic evidence of bleeding was seen, endoscopy was either nondiagnostic (n = 11) or in error (n = 3). We assumed that all patients with anglographic extravasation and upper gastrointestinal bleeding to the extent described herein would require an operation to control hemorrhage. Transcatheter embolization or selective vasopressin infusion averted operation in 13 of these 25 patients (52%). Baum and co-workers [5] and Nusbaum et al [6] established selective mesenteric arteriography as a useful and safe diagnostic tool in patients with upper gastrointestinal hemorrhage. Although early examination by a skilled endoscopist has increased the diagnostic accuracy in patients with this common clinical problem, endoscopy becomes less sensitive and less specific in certain situations, including rapid bleeding and lack of patient cooper-

VOLUME 159

MARCH 1990

ANGIOGRAPHY IN UPPER GASTROINTESTINAL BLEEDING

TABLE V

Need for Operation to Control Massive Upper Gastrointestinal Bleeding Stratified by Diagnostic/Therapeutic Group* Mortality ( % )

Age (yr)

7 7

86 28

7 i 4- 16 49 4- 14

n Group 1 Operation required Operationaverted Group 2 Operationrequired Operation averted Group 3 Operation required Operation averted

5 6

80 50

6 14

Group 4 Operation required Operation averted

6 13

Blood Transfusion (units) Pre-angiogram Post-angiogram

Platelets t (X 103/mm a)

Creatinine t (mg/rnL)

LOS (days)

15 4- 6 8 4- 4

8 -4- 7 8 4- 5

118 4- 73 159 4- 99

2.2 4- 1.7 1.1 4- 0.3

32 ~ 20 10 4- 6

64 4- 4 69 4- 9

6 4- 4 7 4- 3

9 4- 3 5 4- 5

125 4- 30 163 4- 90

1.2 4- 0.1 2.0 4- 1.6

28 4- 19 11 4- 7

66 7

61 4- 8 55 4- 18

12 4- 6 9 J,- 6

12 4- 5 2 4- 2

150 4- 45 204 -I- 109

2.6 4- 1.7 1.4 4- 0.7

40 -4- 44 17 4- 21

33 30

56 4- 15 63 4- 13

10 4- 3 7 4- 4

8 4- 5 4 4- 6

200 4- 100 278 4- 112

3.2 4- 4.4 1.7 4- 5.1

30,4- 15 27 -4- 44

" Group 1 = angiographic extravasation and selective arterial embolization; Group 2 = angiographic extravasation and selective intra-arterial vasopressin; Group 3 --- no extravasation and blind arterial ernbolization; Group 4 = no extravasat)on and no embolization. t On day of anglogram. LOS = length of stay (days until discharge or death).

ation. This is usually due to poor visibility. The poor-risk et al [11] had previously reported the successful control patient with massive upper gastrointestinal bleeding not of duodenal ulcer hemorrhage with autologous clot eminfrequently presents such a challenge to the endoscopist. bolization in a poor-risk patient with gastrointestinal Selective mesenteric angiography remains an attractive bleeding. More recently, Gomes et al [4] described a diagnostic tool in this group. It requires little patient series of 47 patients culled from a larger series of 250 cooperation, and, although gastrointestinal bleeding is patients referred for angiographic evaluation of upper typically intermittent, many patients with massive hem- and lower gastrointestinal bleeding. These 47 patients orrhage (greater than 4 units/24 hours) will be bleeding had arterial extravasation or a vascular abnormality notat a rate fast enough for possible visualization (i.e., great- ed at angiography that was thought to be responsible for their bleeding. They were treated with angiotherapy (eier than 0.5 mL/min). Angiography has also been touted as a useful therapy ther intra-arterial vasopressin or selective embolization), in the poor-risk patient with massive hemorrhage from and successful therapy was defined as angiographic or the gastrointestinal tract. The implication that it is a safer clinical evidence of cessation of bleeding. By these criteor more effective treatment than operation in this group ria, 52% of patients treated with selective vasopressin with prohibitively high operative mortality has not been infusion had their bleeding successfully controlled comproven. Baum and Nusbaum [7] and Athanasoulis et al pared with 88% of patients treated by selective emboliza[8] first showed that selective intra-arterial infusion of tion. They noted that embolization was particularly more vasopressin could safely stop upper gastrointestinal bleed- successful in the treatment of pyloroduodenal bleeding. To convince the experienced surgeon that angiograing from both varices and arteriocapillary sources. Rosch et al [9] first demonstrated that selective arterial emboli- phy is useful in critically ill patients with massive upper zafion could be used to control upper gastrointestinal gastrointestinal bleeding, it must be shown that this mobleedingl Since then, several groups have attempted to dality is helpful diagnostically (i.e,, gives information that corroborate the diagnostic and therapeutic usefulness of makes a subsequent operation better, safer, and easier) or selective mesenteric angiography in massive upper gas- beneficial therapeutically (i.e., improves mortality), or both. Although previous studies suggest that gut angiogtrointestinal bleeding. Athanasoulis et al [8] studied 50 patients with acute raphy is useful in this poor-risk population, many are gastric mucosal hemorrhage who had received an average difficult for the surgeon confronted with such a patient to of 9 units of blood prior to angiography. They found that interpret. Specifically, the definition of successful anintra-arterial infusion of vasopressin controlled bleeding giotherapy is variable and sometimes too lenient. Some in 84% of the 37 patients in whom this was used. This studies do not contain enough information on the rate of contrasted with a control rate of only 3 of 13 patients who bleeding or the degree of illness or risk factors present in were not treated with intra,arterial vasopressin. Success- the study population. Comparisons of endoscopic and ful control of gastric hemorrhage was defined as the pres- angiographic diagnoses are inconstant. Finally, some ence of the following for 5 days after vasopressin infusion: studies do not report the need for subsequent operation to non-bloody nasogastric tube aspirate, no further need for control hemorrhage after an attempt at angiographic blood transfusion, and stable hematocrit. Eisenberg and control, and many do not give mortality rates. We have tried to correct many of these deficiencies in Steer [10] reported an overall success rate of 50% in stopping pyloroduodenal bleeding with angiographic au- the current study and conclude that gut angiography has tologous clot embolization in six poor-risk patients. White diagnostic and therapeutic potential in the poor-risk paTHE AMERICAN JOURNAL OF SURGERY

VOLUME 159

MARCH 1990

285

DEMPSEY ET AL

tient with massive nonvariceal upper gastrointestinal bleeding. This procedure shows the site of bleeding in approximately 40% of this patient group and is often positive (22%) despite a nonconclusive or even erroneous upper endoscopy. Based upon the site of extravasation, angiographically directed therapy (selective embolization or selective vasopressin infusion) m a y avert operation in up to 52% of this patient subgroup. If operation can be avoided in these patients with angiographic extravasation, mortality is halved (83% versus 38%). The majority of patients with arteriocapillary upper gastrointestinal bleeding in whom contrast extravasation is not shown on angiogram will stop bleeding without an operation. In this group, blind transcatheter embolization does not appear to be helpful in avoiding an operation when compared with continued supportive care. Based upon our findings and those of others, we believe that gut angiography is a safe and effective procedure that should be applied selectively to patients with massive upper gastrointestinal bleeding. We use this modality most often in poor-risk patients with questionable or nondiagnostic findings on upper endoscopy and in patients unlikely to tolerate a major operation. Most candidates in our practice come from the intensive care unit and have multiple system failure. If contrast extravasation is demonstrated, angiotherapy (usually directed arterial embolization) is tried. If bleeding is not controlled, the patient is taken to the operating room immediately. We are not optimistic regarding the efficacy of blind embolization in patients without an angiographically documented bleeding site. However, we continue to use this technique in the occasional poor-risk patient because of its apparent safety.

This study supports the use of angiography in massive upper gastrointestinal bleeding to enhance diagno-

286

sis, which it did in 22% of patients, and to stop bleeding, which it did in 52% of patients who had a lesion diagnosed by angiography and in whom angiotherapy was tried.

REFERENCES 1. Kim B, Wright HK, Bordan D, Fielding LP, Swaney R. Risks of surgery for upper gastrointestinal hemorrhage: 1972 vs. 1982. Am J Surg 1985; 149: 474-6. 2. Larson G, Schmidt T, Gott J, Bond S, O'Connor CA, Richardson JD. Upper gastrointestinal bleeding: predictors of outcome. Surgery 1986; 100: 765-72. 3. Kelemouridis V, Athanasoulis CA, Waltman AC. Gastric bleeding sites: an angiographic study. Radiology 1983; 149: 643-8. 4. Gomes AS, Lois JF, McCoy RD. Angiographic treatment of gastrointestinal hemorrhage: comparison of vasopressin infusion and embolization. AJR 1986; 146: 103l-7. 5. Baum S, Nusbaum M, Blakemore WS, Finkelstein A. The preoperative radiographic demonstration of intra-abdominal bleeding from undetermined sites by percutaneous selective celiac and superior mesenteric arteriography. Surgery 1965; 58: 797-805. 6. Nusbaum M, Baum S, Blakemore WS. Clinical experience with the diagnosis and management of gastrointestinal hemorrhage by selective mesenteric catheterization. Ann Surg 1969; 170: 506-14. 7. Baum S, Nusbaum M. The control of gastrointestinal hemorrhage by selective mesenteric arterial infusion of vasopressin. Radiology 1971; 98: 497-505. 8. Athanasoulis CA, Baum S, Waltman AC, et al. Control of acute gastric mucosal hemorrhage: intra-arterial infusion of posterior pituitary extract. N Engl J Med 1974; 290: 597-603. 9. Rosch J, Dotter CT, Brown MJ. Selective arterial embolization: a new method for control of acute gastrointestinal bleeding. Radiology 1972; 102: 303-6. 10. Eisenberg H, Steer ML. The nonoperative treatment of massive pyloroduodenal hemorrhage by retracted autologous clot embolization. Surgery 1976; 79: 414-20. 11. White RI Jr, Giargiana FA, Bell W. Bleeding duodenal ulcer control: selective arterial embolization with autologous blood clot. JAMA 1974; 229: 546-8.

THE AMERICAN JOURNAL OF SURGERY VOLUME 159 MARCH 1990