When to discharge patients with bleeding peptic ulcers: a prospective study of residual risk of rebleeding

0016-5107/96/4404-038255.00 + 0 GASTROINTESTINAL ENDOSCOPY Copyright © 1996 by the American Society for Gastrointestinal Endoscopy

When to discharge patients with bleeding peptic ulcers: a prospective study of residual risk of rebleeding Ping-I Hsu, MD, Kwok-Hung Lai, MD, Xi-Zhang Lin, MD, Yun-Fu Yang, MD, Mike Lin, MD Jeng-Shiann Shin, MD, Gin-Ho Lo, MD, Rong-Long Huang, MD, Chia-Fu Chang, MD Chiun-Ku Lin, MD, Luo-Ping Ger, MPH Kaohsiung and Tainan, Republic of China

Background: From January 1993 to December 1994, we conducted a prospective study to investigate the evolutionary change of rebleeding risk in bleeding peptic ulcers. To obviate possible confounding factors that would influence decision making for discharge of patients, subjects with coexistent acute illnesses, systemic bleeding disorders, alcoholism, and use of nonsteroidal anti-inflammatory drugs were excluded. Methods: Emergency endoscopies were performed in patients with hematemesis or a melena within 24 hours of admission. Ulcer lesions were divided into six categories according to endoscopic findings. The residual risks of rebleeding of each type of ulcers were calculated for 10 days, and the critical point of acceptable rebleeding risk after discharge was set at 3%. Results: Three hundred ninety-two patients with bleeding peptic ulcers completed the study. The ulcers, characterized by clean bases, red or black spots, adherent clots, nonbleeding visible vessels without local therapy, nonbieeding visible vessels with local therapy, and bleeding visible vessels with local therapy took 0, 3, 3, 4, 4, and 3 days, respectively, to decrease rebleeding risk to below the critical point. All episodes of fatal rebleeding (n = 4) occurred within 24 hours after admission. Conclusions: Patients with clean-based ulcers can be discharged in the first day of admission. The optimal duration required for hospitalization of patients with ulcers characterized by nonbleeding visible vessels at initial endoscopy is 4 days. The remaining patients with ulcers marked by other bleeding stigmata may be discharged after a 3-day observation. (Gastrointest Endosc 1996;44:382-7.)

Bleeding is a common but serious complication of peptic ulcer disease. At least 80% of bleeding episodes from peptic ulcers cease spontaneously. 1, 2 For these Received August 15, 1995. For revision September 22, 1995. Accepted January 15, 1996. From the Department of Emergency, Division of Gastroenterology, the Department of Medicine, and the Department of Education and Research, Veterans General Hospital-Kaohsiung; and the Division of Gastroenterology, Department of Internal Medicine, National Cheng Kung University Hospital, Taiwan, R.O.C. Reprint requests: Kwok-Hung Lai, MD, Gastroenterology, Veterans General Hospital-Kaohsiung, 386, Ta-Chung 1st Road, Kaohsiung 813, Taiwan, Republic of China. 37/1/72068

382

GASTROINTESTINAL ENDOSCOPY

patients, the treatment course is comparatively uncomplicated and the mortality rate is 4% or lessfi 4 In the remaining patients with persistent or recurrent bleeding, the death rate is 12% to 18%. 5-7 Therefore, identification and hospitalization of patients with a high risk of further bleeding is quite important. However, in clinical practice, determining the duration of hospital stay for patients with bleeding ulcers is another crucial problem for physicians. A prolonged stay in hospital increases medical costs, while inappropriately early discharge may result in an increase in preventable death of patients with severe rebleeding. Our previous study demonstrated that the stigmata of hemorrhage was a very important predictor of VOLUME 44, NO. 4, 1996

rebleeding. 6 A l t h o u g h m o s t studies 3, 6-10show t h a t t h e m a j o r i t y of episodes of r e c u r r e n t bleeding occur w i t h i n 48 to 96 h o u r s a f t e r initial bleeding, t h e r e h a s b e e n no prospective s t u d y i n v e s t i g a t i n g the r e s i d u a l r i s k of r e b l e e d i n g b a s e d on the a p p e a r a n c e of s t i g m a t a of recent h e m o r r h a g e (SRH). I n addition, m o s t r e p o r t s t h a t discussed t h e t i m e l a p s e of r e b l e e d i n g w e r e b a s e d on t h e n a t u r a l h i s t o r y of bleeding ulcers. 2, 9, lo Time-dep e n d e n t studies concerning r e s i d u a l r i s k o f r e b l e e d i n g on each successive d a y of peptic ulcers following t h e r a p e u t i c endoscopy a r e rare. T h e p u r p o s e of this s t u d y is to i n v e s t i g a t e the evolution of r e b l e e d i n g r i s k in peptic ulcers c h a r a c t e r i z e d b y different t y p e s of S R H on initial endoscopy a n d to develop a s t r a t e g y for d i s c h a r g i n g p a t i e n t s w i t h bleeding ulcers. To obviate possible confounding factors t h a t would influence decision m a k i n g for discharge of p a t i e n t s , subjects w i t h coexistent a c u t e illnesses, syst e m i c bleeding disorders, alcoholism, a n d u s e of nonsteroidal a n t i - i n f l a m m a t o r y d r u g s w e r e excluded. PATIENTS AND METHODS Patients

From J a n u a r y 1993 to December 1994, 712 consecutive patients with hematemesis, melena, or both, had emergency upper endoscopy performed within 24 hours of admission to the emergency units of Veterans General Hospital-Kaohsiung and the National Cheng Kung University Hospital. All patients with bleeding peptic ulcers proven by endoscopy who gave their consent were enrolled in the study. Criteria for exclusion included (1) the presence of other possible bleeding sites (for example, esophageal varices, gastric cancer), (2) the intake ofnonsteroidal anti-inflammatory drugs within 1 week before admission, (3) the coexistence of an acute significant illness (for example, sepsis, stroke, acute myocardial infarction, acute respiratory failure, acute renal failure, acute hepatic failure, acute surgical abdomen) or a malignancy, (4) alcoholism (->80 gin/day alcohol consumption for at least 5 years), (5) the presence of a systemic bleeding tendency (for example, platelet count -<50,000/ m m 3, decompensated liver cirrhosis with prolonged prothrombin time >-4 seconds, disseminated intravascular coagulopathy, or use of an anticoagulant). Since the aim of our study was to determine the optimal duration required for hospitalization of patients with bleeding ulcers, we excluded patients with other conditions which might prolong hospitalization. Methods

Emergency endoscopies were performed within 24 hours of admission in all patients. The equipment used was the Olympus GIF XV10, the GIF XQ 200, and the GIF 1T20 (Olympus Corp., Tokyo, Japan). To improve the visual field, gastric ]avage was carried out before endoscopy. Ulcers with stigmata were cleaned by water irrigation through the biopsy channel. Adherent clots were not removed with biopsy forceps. We divided the ulcer lesions into six categories according to a modified Wara's classification7: (1) clean VOLUME 44, NO. 4, 1996

base, (2) red or black spot, (3) oozing (without visible vessel), (4) adherent clot, (5) nonbleeding visible vessel (NBVV), and (6) bleeding visible vessel (either spurting or oozing) (BVV). An NBVV was defined as a raised red or bluish-red hemispheric lesion protruding from the ulcer base, without active bleeding. An adherent clot was defined as an overlying clot that was resistant to washing. A red or black spot was defined as a localized, small, red or black stigmata that was not protruding from the ulcer base. "Active bleeding" was defined as the presence of either a "bleeding visible vessel" or "oozing (without a visible vessel) from the ulcer base or margin." If active bleeding was noted during the first examination, endoscopic local injection with diluted epinephrine solution (0.01% epinephrine, 0.5 to 1 mL per injection) was performed for hemostasis. Local injection was performed at four quadrants, 2 mm away from the bleeder. Initial hemostasis was defined as no bleeding from the ulcer for at least 5 minutes. If bleeding persisted even after the total injected volume of diluted epinephrine exceeded 10 mL, then failure was conceded. Patients who failed to achieve hemostasis received heater probe thermocoagulation or surgical intervention. For every patient with NBVV, a sealed envelope was opened to decide the treatment modalities during the emergency endoscopic procedure, the contents of the serial envelopes having been previously randomized by a statistician (Luo-Ping Get, MPH) according to "observation" or "local injection." Patients with other stigmata of recent hemorrhage or clean-based ulcers were observed only. To assess the significance of clinical and laboratory factors in predicting rebleeding, the following data were recorded for each patient: age, sex, past history of upper gastrointestinal bleeding, history of smoking, coexistence of an underlying medical disease (including diabetes mellitus, hypertension, significant heart disease, chronic lung disease, and liver cirrhosis without prolonged prothrombin time), initial blood pressure, initial pulse rate, presence of hypovolemic shock before endoscopy, and serum blood urea nitrogen, creatinine, and alanine aminotransferase levels on admission. Shock was defined as systolic blood pressure less than 90 mm Hg and a pulse rate greater than 100 beats/minute, accompanied by pallor, cold sweats, and oliguria. Patients who were hemodynamically stable on admission and had clean ulcer bases as shown by initial endoscopy were discharged within 24 hours. Patients who had active bleeding or SRH were admitted for close observation. During the stay in the hospital, patients were given partial parenteral nutrition and intravenous H2 blockers. Oral intake was restricted to medication necessary to control other medical diseases. After a 48-hour observation, patients were given a clear liquid diet for 24 hours, and then a regular diet. The hemoglobin level was checked every day, and a blood transfusion was given if the hemoglobin concentration dropped to less than 8 gln/dL or if vital signs deteriorated. Assuming that there was no other reason for hospitalization, they were discharged 4 days after admission. The patients who developed rebleeding during the observation period were initially treated with either endoscopic injection therapy or heater probe thermocoagulation. If hemostasis could not be achieved, surgical intervention was conducted. After discharge, patients received oral H2 blockers and were followed-up by telephone every other day. Patients GASTROINTESTINAL ENDOSCOPY

383

25

w

•=

-o-Clean base -c-Spot -*-Clot

==Is

¢¢ =~I0 w.Q

0 0

1

2

3

4

5

6

7

8

9

10

Days Figure 1. Evolution of residual risk of rebleeding of ulcers characterized by clean bases (N = 139), red or black spots (N = 81), and adherent clot (N = 64) on each successive day after admission.

were also requested to return to the outpatient clinic on day 3 and day 7 to check for evidence of rebleeding. The end points of this study were (1) death of the patient, or (2) observation for 10 days. A clinician who was independent ofthe endoscopist observed the patient in the hospital for evidence of rebleeding. Evidence of rebleeding was regarded as deftnite if fresh hematemesis occurred with a pulse rate greater than 100 beats/minute, a fall in systolic blood pressure exceeding 30 mm Hg, or a decrease in hemoglobin of at least 2 gm/dL after initial stabilization, but as suggestive if there was fresh melena, a sudden rise in pulse rate, or a fall in blood pressure only. In case Of suggestive evidence, rebleeding was regarded as definite only if it was confirmed by repeat endoscopy. Statistics

We used the chi-squared test with or without Yate's correction to compare the rebleeding rates in the 14 clinical and endoscopic variables. These 14 variables were as follows: age (<60 or ->60 years), gender, past history of peptic ulcer bleeding, history of smoking, coexistence of an underlying medical disease, initial pulse rate (---100 beats/ minute or >100 beats/minute), presence of hypovolemic shock before endoscopy, initial hemoglobin (<10 or ->10 gm/ dL), location of ulcer (stomach, duodenum), blood urea nitrogen (-<19 or >19 mg/dL), blood creatinine (-<1.4 mg/dL or >1.4 mg/dL), ulcer size (<2.0 or ->2.0 cm), number of ulcers (1 or >1), and type of ulcer (clean base, red or black spot, adherent clot, oozing, NBVV without local therapy, NBVV with local therapy, BVV with local therapy). Ap value less than 0.05 was considered to be significant. Those variables found to be significant by univariate analysis were subsequently assessed by a stepwise logistic regression to identify those that were independently significant in predicting rebleeding. The "residual risk" of rebleeding was determined by calculating the resulting risk of rebleeding after a specific number of days had passed since admission. The critical point of acceptable rebleeding risk following discharge was discussed by experienced endoscopists (including Drs. Ping-I Hsu, Kwok-Hung Lai, Xi-Zhang Lin, GinHo Lo, Jeng-Shiann Shin) in the clinical research commit384 G A S T R O I N T E S T I N A L E N D O S C O P Y

tee of the Veterans General Hospital-Kaohsiung, and Was set at 3%. RESULTS

There were 473 patients with bleeding peptic ulcers proven by endoscopy in the study period. Sixty-nine patients were excluded for the following reasons: 16 patients had other potential sources of bleeding (15 with esophageal varices, i with gastric cancer); 22 took nonsteroidal anti-inflammatory drugs for ischemic h e a r t disease, arthritis, or cerebral infarction within 1 week before admission; 26 patients had acute significant illnesses; 2 patients were alcoholics; and 3 patients had systemic bleeding disorders. In the excluded group, 2 patients with acute illnesses and one taking nonsteroidal anti-inflammatory drug rebled. Three of the 26 patients with acute illnesses died of nonulcer, underlying diseases (1 each of septic shock abdominal trauma, and head injury). Four hundred four patients with peptic ulcer bleeding were initially enrolled in the study. Twelve (7 with clean bases, 3 with fiat red or black spots, 1 with adherent clot, and 1 with NBVV) were lost to follow-up after discharge and were excluded from the study. None of t hem rebled during hospitalization. Of the remaining 392 patients with peptic ulcers, 26 had BVVs, 74 had NBVVs, 64 had adherent clots, 81 had flat red or black spots, 8 had oozing without visible vessels at the ulcer bases, and 139 had clean ulcer bases. The age of these patients was 59 + 16 (mean +_ SD, range; 15 to 93) years, and the male to female ratio was 309/83. There were 180 patients with gastric ulcer, and 212 with duodenal ulcers. Ulcers with clean base

In the 139 patients with clean-based ulcer, 2 patients (1.4%) rebled. One episode occurred on day 3, and the other occurred on day 7. Oozing from the ulcer base was found on repeated endoscopy at the time of rebleeding. Both patients were successfully treated by endoscopic injection therapy. The overall death rate was zero. Residual risk of rebleeding in this patient group is shown in Figure 1. The rebleeding risk was less t h a n 3% on day 0. Ulcers with a red or black spot

Among the 81 patients with a flat red or black spot on initial endoscopy, 3 (3.7%) rebled. Of these patients, one ceased bleeding with supportive care only. The other two were successfully treated with therapeutic endoscopy (one with epinephrine injection and the other with heat er probe thermocoagulation). The overall death rate was zero. Residual risk of rebleeding in this patient group is shown in Figure 1, and the rebleeding risk was less t h a n 3% after a 3-day observation. VOLUME 44, NO. 4, 1996

Ulcers with oozing from ulcer base In eight patients with oozing from ulcer bases on initial endoscopy, one patient (12.5%) rebled. The rebleeding episode occurred on day 4, and the oozing ulcer was then successfully treated by heater probe thermocoagulation. There was no mortality in this patient group. The residual risk ofrebleeding was not analyzed because there were too few cases.

25

I

" 6 ~ 20

BW+LT .-g__~ 10 me,,

=

0

1

2

3

4

Ulcers with an adherent clot In the 64 patients with adherent clots in ulcer bases, 9 patients rebled. The time lapse ofrebleeding ranged from 1 to 5 days. Most of these patients (77.8%) had BBVs present at the time of rebleeding. Hemostasis was achieved by epinephrine injection in 2 patients, by heater probe thermocoagulation in 2 patients, and by supportive care only in 1 patient. Four patients required emergent operation for uncontrolled rebleeding, and 2 died of surgical complications (both complicated with sepsis and adult respiratory distress syndrome). Both the fatal rebleeding episodes occurred within 24 hours after admission. The overall death rate was 3.1%. Residual risk of rebleeding in this patient group is shown in Figure 1, and the rebleeding risk was less than 3% after a 3-day observation. Ulcers with NBVVs and without local therapy Among the 36 patients with N B W s and without injection therapy, 8 patients (22.2%) rebled. Five of the 8 rebleeders received successful endoscopic treatment (3 with epinephrine injection and 2 with heater probe thermocoagulation). Hemostasis was not achieved in 3 patients. They all subsequently received surgical intervention. The overall death rate in this patient group is zero. Residual risk ofrebleeding in the patient group was shown in Figure 2, and the rebleeding risk was less than 3% after a 4-day observation. Ulcers with NBVVs and local therapy In the 38 patients with NBVVs and epinephrine injection therapy, 5 patients (13.2%) rebled. The time lapse of rebleeding ranged from 2 to 5 days. Ultimate hemostasis was achieved by epinephrine injection in I patient, by heater probe thermocoagulation in 1 patient, and by supportive care only in 2 patients. The other 1 required surgical intervention. The death rate was zero in this patient group. Residual risk of rebleeding is shown in Figure 2, and the rebleeding risk was less than 3% after a 4-day observation. Among patients characterized by NBVVs, local injection with epinephrine reduced rebleeding rate from 22.2% to 13.2%. However, the difference ofrebleeding rates between groups with and without epinephrine injection did not reach statistical significance (p > 0.05). VOLUME 44, NO. 4, 1996

--o- NBW --c--NBW+LT

5 6 Days

7

8

9

10

Figure 2. Evolution of residual risk of rebleeding of ulcers with nonbleeding visible vessels and bleeding visible vessels. NBVV, Nonbleeding visible vessels without local therapy (N = 36); NBVV+ LT, nonbleeding visible vessels with local therapy (N = 38); BVV+ LT, bleeding visible vessels with local therapy (N = 24).

Patient group with BVVs and local therapy Twenty-six patients with BVVs were treated by epinephrine injection. Hemostasis was not achieved in 3 patients. One of them then received successful heater probe thermocoagulation. Another 1 had emergency surgery, but died from surgical complications (sepsis with acute respiratory distress syndrome). The other one died of continuous bleeding before surgical intervention. Of the 24 patients with BVVs that were controlled temporarily by therapeutic endoscopy, 4 (16.7%) suffered from rebleeding during the follow-up period. Two of the 4 rebleeders received heater probe thermocoagulation for ultimate hemostasis, and 2 required surgical intervention. The overall death rate was 7.7% (2 of 26). Figure 2 shows the residual risk of rebleeding in this group of patients, and the rebleeding risk was less than 3% after a 3-day observation. Risks factors of rebleeding In the 390 patients without active bleeding at initial endoscopy or with active bleeding that was successfully controlled by local therapy, univariate analysis of the 14 clinical and endoscopic variables showed six factors significantly associated with rebleeding: past history of peptic ulcer bleeding (p < 0.01), history of smoking (p < 0.05), presentation with hematemesis (p<0.01), hypovolemic shock before endoscopy (p < 0.001), ulcer size ->2.0 cm (p < 0.01), and type of ulcer (p < 0.001). The above six factors were then entered in the stepwise logistic regression. Multivariate analysis with stepwise logistic regression showed that a past history of peptic ulcer bleeding, hypovolemic shock, ulcer size 2.0 cm or greater, adherent clot, NBVV without local therapy, NBVV with local therapy, and B W with local therapy were independently significant in predicting rebleeding (p < 0.05) (Table 1). G A S T R O I N T E S T I N A L E N D O S C O P Y 385

Table 1. Logistic regression of risk factors for rebleeding Variable

Coefficient

SE

Relative risk (95% CI*)

•43 3.22 (1.37-7.59) History of peptic ulcer bleeding: + vs 1.17 •53 3.14 (1.12-8.84) Hypovolemicshock: + vs 1.14 •51 4.07 (1.51-10.99) Ulcer size: >-2 cm vs <2 cm 1.40 Ulcer type* •82 8.82 (1.77-44.05) Adherent clot vs clean base 2.18 .84 14.22(2.72-74.44) NBVVwithout therapy vs clean base 2.65 •88 10.67(1.91-59.50) NBW with therapy vs clean base 2.37 .93 11.70(1.90-72.24) BVV with therapy vs clean base 2.46 CI, Confidenceinterval. *Ulcers with oozing from base were not analyzedbecause there were too few cases.

DISCUSSION Determining the optimal duration of hospitalization for patients with bleeding ulcers is an important problem for physicians. Some authors 11 proposed t h a t the disappearance of SRH may be used as a valuable criteria for discharge of ulcer patients. However, it is not practical to perform repeated endoscopy on the patients every day to monitor the evolution of bleeding stigmata in clinical practice because of both the poor compliance of patients and the tremendous medical costs. Our previous study showed t h a t observation for 96 hours is sufficient to detect most rebleeding episodes after an initial bleeding from peptic ulcer. 6 However, a 4-day hospitalization after ulcer hemorrhage seems unnecessary for patients with ulcers characterized by clean base or old stigmata on initial endoscopy. 12-14 To establish the best and most cost-effective management for patients with bleeding ulcer, we analyzed the residual risk of rebleeding for each type of ulcer lesions. According to our study, if a 3% rebleeding risk is an acceptable critical point for discharge, the optimal duration of hospitalization of the patients with bleeding ulcers characterized by clean bases, red or black spots, adherent clots, NBVVs without local therapy, NBVVs with local therapy, and BVVs with local therapy are 0, 3, 3, 4, 4, and 3 days, respectively. It is important to point out t h a t the cutoffvalue of acceptable rebleeding risk may depend on the experiences of clinicians and the medical environment in their countries. In clinical practices, they can decide when to discharge their patients according to their acceptable rebleeding risk and our illustrations for residual risk of rebleeding on each successive day. However, we would emphasize that the results currently apply only to patients without coexistent acute illnesses, alcoholism, use of nonsteroidal anti-inflammatory drugs, or bleeding disorders inasmuch as we excluded these patients in the study. In general, our results support the algorithm proposed by Laine and Peterson, 15 who suggested that patients with clean-based ulcers may be sent directly home and patients with bleeding ulcers characterized 386 G A S T R O I N T E S T I N A L E N D O S C O P Y

p-value 0.0074 0.0300 0.0055 0.0080 0.0017 0.0070 0.0081

by any SRH may be considered for discharge after 3 days. It is important to point out t h a t all rebleeding episodes associated with mortality in this series happened in patients with adherent clots or BVVs at initial endoscopy and all occurred within 24 hours after admission. According to our strategies and those of Laine and Petersen, no life-threatening rebleeding was missed. In our study and most other reports, the rate of recurrent bleeding in patients with clean-based ulcers ranged from 0% to 2%. 12-15Therefore, the residual risk of rebleeding was less t h a n 3% even on day 1. Such patients are candidates for early discharge (i.e., during the first day) after stabilization of vital signs. Among patients with ulcers marked by black or red spots, the incidence of rebleeding ranged from 0% to 13%.7, 13,14-18 The evolution of residual risk in this patient group has not been well investigated before. Our study showed t h a t such patients still required a 3-day hospitalization to reduce the rebleeding risk to less t h a n 3%. Patients with bleeding ulcers characterized by adherent clots faced a high rebleeding risk on admission.7, 10, 13, 16-19 However, the risk decreased day by day after admission, and dropped to less t h a n 3% after a 3-day observation in our study. Our result was corroborated by another report in which follow-up endoscopy was performed daily in patients who had bleeding ulcers with sentinel clots. The stigmata disappeared within 3 days in all the patients. 7 In our series, the rebleeding rate of ulcers with NBVVs was 22.2%, which is fairly consistent with most other reports. 7, 9, 12, 13, 17, 2o, 21 Although endoscopic injection has been shown to be effective to arrest peptic ulcer hemorrhage 5, 20 and the technique did reduce rebleeding risk from 22.2% to 13.2% in this prospective, randomized controlled triM, the difference in rebleeding rates between patient groups with and without epinephrine injection did not reach statistical significance. The relatively small case number may be a problem for comparison in this clinical trial. Another possible explanation for the unremarkable effect of VOLUME 44, NO. 4, 1996

therapeutic endoscopy may lie in the fact that some patients with high risk for rebleeding were presumably excluded from the study on the basis of our exclusion criteria (e.g., the coexistence of an acute significant illness, the intake ofnonsteroidal anti-inflammatory drugs). They may get more benefits from therapeutic endoscopy. It is not clear whether the observation duration required for ulcers characterized by NBVVs is longer than the ones for other types of lesions. However, our current results are consistent with one of our previous studies, which demonstrated that a visible vessel takes 4.1 + 2.1 days to disappear, requiring significantly more time than an adherent clot or a red or black spot, which take 2.4 _+ 0.8 days and 2.4 _+ 1.3 days, respectively. 11 Using multivariate analysis, we demonstrated that a past history of peptic ulcer bleeding, hypovolemic shock, and an ulcer size of 2.0 cm or greater were also independently significant in predicting rebleeding. It seems that a combination of clinical and endoscopic factors may provide a more reliable prediction of rebleeding than a single criteria alone. For example, an adherent clot associated with hypovolemic shock had a rebleeding rate of 37.5%, but the presence of a clot without hypovolemic shock had a rebleeding rate of only 10.7%. The clinical significance of hypovolemic shock is also present in patients with old SRH. In our series, a red or black spot associated with hypovolemic shock had 50% (2 of 4) rebleeding rate, while a red or black spot without hypovolemic shock had a rebleeding rate of only 1.3% (1 of 77). It is possible that a strategy considering all independent predictors of rebleeding may be more valuable in decision making for discharge of patients and deserves further investigation in the future. In summary, we developed a scheme according to the residual risk of rebleeding for discharge of patients with bleeding ulcers who are otherwise healthy and do not take nonsteroidal anti-inflammatory drugs. Patients with clean-based ulcers can be discharged on the first day of admission, assuming that their vital signs have been stabilized. The optimal duration required for hospitalization of the patients whose ulcers are characterized by nonbleeding visible vessels is 4 days. The remaining patients with ulcers characterized by other stigmata of recent hemorrhage may be discharged after a 3-day observation if there is no other reason for hospitalization. ACKNOWLEDGMENTS The authors express their appreciation to Drs. JinShiung Cheng, Sam-Ming Chen, Shi-Chi Wen, Jia-

VOLUME 44, NO. 4, 1996

Shen Huang, and Jyh-Wei Chen for their invaluable support in this study, and to Dr. Susan Shin-Jung Lee for revising the manuscript. REFERENCES 1. Silverstein FE, Feld AD, Gilbert DA. Upper gastrointestinal tract bleeding. Arch Intern Med 1981;141:322-7. 2. Northfield TC. Factors predisposing to recurrent haemorrhage after acute gastrointestinal bleeding. BMJ 1971;1:26-8. 3. Fleischer D. Etiology and prevalence of severe persistent upper gastrointestinal bleeding. Gastroenterology 1983;84:73-9. 4. Schiller KFR, Truelove SC, William DG. Haematemesis and melena, with special reference to factors influencing the outcome. BMJ 1970;2:7-14. 5. Lin HJ, Chan FY, Lee ZC, et al. Endoscopic injection to arrest peptic ulcer hemorrhage: a prospective, randomized controlled trial; preliminary results. Hepatogastroenterology 1991;38: 291-4. 6. Hsu PI, Lin XZ, Chan SH, et al. Bleeding peptic ulcer--risk factors for rebleeding and sequential changes in endoscopic findings. Gut 1994;35:746-9. 7. Wara P. Endoscopic prediction of major rebleeding--a prospective study of stigmata of hemorrhage in bleeding ulcer. Gastroenterology 1985;88:1209-14. 8. DeDombal FT, Clarke JR, Clamp SE, Malizia G, Kotwal MR, Morgan AG. Prognostic factors in upper GI bleeding. Endoscopy 1986;18:$6-10. 9. Lin HJ, Perng CL, Lee FY, Lee CH, Lee SD. Clinicalcourses and predictors for rebleeding in patients with peptic ulcers and non-bleedingvisible vessels: a prospective study. Gut 1994;35: 1389-93. 10. Chung SCS, Leung JWC, Lo KK, So LYS, Li AKC. Natural history of the sentinel clot: an endoscopic study [abstract]. Gastroenterology 1990;98(suppl):A31. 11. Yang CC, Shin JS, Lin XZ, Hsu PI, Chen KW, Lin CY. The natural history (fading time) of stigmata of recent hemorrhage in peptic ulcer disease. Gastrointest Endosc 1994;40:562-6. 12. Emmanuel A, Rokkas T. Prognostic importance of visible vessels in hemorrhage of peptic ulcers [letter]. Gastrointest Endosc 1985;31:52. 13. Matthewson K, Swain CP, Bland M, Kirkham JS, Bown SG, Northfield TC. Randomized comparison of Nd:YAG laser, heater probe, and no endoscopic therapy for bleeding peptic ulcers. Gastroenterology 1990;98:1239-44. 14. Laine L, Cohen H, Brodhead J, Cantor D, Garcia F, Mosquera M. Prospective evaluation of immediate versus delayed refeeding and prognostic value of endoscopy in patients with upper gastrointestinal hemorrhage. Gastroenterology 1992;102: 314-6. 15. Laine L, Petersen WL. Bleeding peptic ulcer. N Engl J Med 1994;331:717-27. 16. Chang-Chien CS, Wu CS, Chen PC, et al. Different implications of stigmata of recent hemorrhage in gastric and duodenal ulcers. Dig Dis Sci 1988;33:400-4. 17. Brearley S, Morris DL, Hawker PC, Dykes PW, Keighley MRB. Prediction of mortality at endoscopy in bleeding peptic ulcer disease. Endoscopy 1985;17:173-4. 18. Storey DW, Bown SG, Swain CP, Salmon PR, Kirkham JS, Northfield TC. Endoscopic prediction of recurrent bleeding in peptic ulcers. N Engl J Med 1981;305:915-6. 19. Krejs GJ, Little KH, Westergaard H, Hamilton JK, Spady DK, Polter DE. Laser photocoagnlation for the treatment of acute peptic-ulcer bleeding: a randomized controlled clinical trial. N Engl J Med 1987;316:1618-21. 20. Lin HJ, Lee FY, Kang WM, Tsai YT, Lee SD, Lee CH. A controlled study of therapeutic endoscopy for peptic ulcer with non-bleedingvisiblevessel. Gastrointest Endosc 1990;36:241-6. 21. Johnston JH. Endoscopic risk factors for bleeding peptic ulcer. Gastrointest Endosc 1990;36:$16-20.

G A S T R O I N T E S T I N A L E N D O S C O P Y 387