Endoscopic Prediction of Major Rebleeding-A Prospective Study of Stigmata of Hemorrhage in Bleeding Ulcer

GASTROENTEROLOGY 1985;88:1209-14 Endoscopic Prediction of Major Rebleeding-A Prospective Study of Stigmata of Hemorrhage in Bleeding Ulcer pal WARA Surgical Gastroenterological Department, Aarhus Municipal Hospital, University of Aarhus, Aarhus, Denmark Two hundred fifty patients with stigmata of active or recent hemorrhage from peptic ulcer were studied. Stigmata, subgrollped according to bleeding status and the presence or absence of a visible vessel, were related to the subsequent clinical course of hemorrhage. Sixty patients (24%) rebled massively and required emergency hemostasis. At the initial endoscopy a visible vessel was a relatively rare finding (21 %). Less than one-third of the visible vessel~ rebled massively. None of the stigmata or subgroups of stigmata emerged as reliable predictors of m~jor rebleeding. Blf)eding status and uker site, however, were observed to influence the relative importance of a visible vessel. When a visible vessel was identified in patients with oozing, overlying clot, or gastric or duodenal ulcer, the probability of major rebleeding increased, but not significantly. However, when a visible vessel was identified in patients without other stigmata or in patients with prepyloric ulcer or older stigmata, there was a decreased probability of major rebleeding. Older stigmata was a superior predictor of self-limited hemorrhage. Stigmata of hemorrhage observed at emergency endosc'1py have been reported to provide important prognostic information. Controversy, however, exists with regard to the significance of the different stigmata in predicting major hemorrhage. Focusing on ulcer bleeding, the aim of the present study was to explore whether stigmata of active or recent hemorrhage are valuable in predicting major rebleeding in need of emergellcy surgery. Material aml Methods To identify all patients with ulcer bleeding, eil1ergency endoscopy was performed in a consecutive series of Received January 10, 19134. Accepted November 30, 1984. Address requests for reprints to: Pill Wara, M.D., Skrering Hedevej 178, DK-8250 Aarhus/Egaa, Denmark. 1985 by the AIl1erical1 Gastroenterological Associqtion 0016-5085/85/$3.30 patients admitted because of hematemesis and melena within the preceding 24 h. A hemodynamic assessment distinguished between self-limited hemorrhage and major hemorrhage requiring emergency surgery. Patients with stable circulation on admission and not in need of emergency control of hemorrhage were observed and monitored for 5 days in a surgical intensive care unit for assessment of further hemorrhage. Major rebleeding was defined as recurrent bleeding accompanied by a sudden decrease of central venous pressure or systolic blood pressure exceeding 5 cmh 2 0 or 50 mmhg, respectively, or bleeding requiring more than 7 U of blood within 48 h. No further hemorrhage after admission and rebleeding accompanied by only insignificant change of the hemodynamic variables were classified as self-limited hemorrhage requi~ing no hemostatic treatment. During 3.5 yr, 276 of a total of 539 patients were admitted with a bleeding peptic ulcer. The median age was 68 yr with a female to male ratio of 1: 1.8. Two hundred seventy-three patients underwent emergency endoscopy performed within 6 h in 94% of the cases (median value 2 h). Gastric lavage, using a double-lumen 36F tube, was carried put if the view was inadequate. Using a treatment endoscope with a 3.7-mm channel for adeqllate suction, the stigmatized ulcer was washed and cleared of blood and nonadhering clots. An ulcer was accepted as the bleeding source only if marked by stigmata of active or recent hemorrhage. Stigmata were subgrouped according to bleeding status (active or recent hemorrhage) and the presence or absence of a visible vessel. Active bleeding was defined as (a) arterial jet, (b) oozing 4emorrhage, or (c) oozing beneath an overlying clot. Recent hemorrhage was indicated by (a) an overlying clot adhering to the ulcer base but without oozing, (b) older stigmata defined as altered blood appearing as black slough adhering to the ulcer base, or (c) bare visible vessel without other stigmata. A visible vessel, bare or together with other stigmata, was defined as an elevated red or bluish-red hemispheric plug protruding from the ulcer base and resistant to washing, a pulsatile pseudoaneurysm, or a Yellow:white rod sticking out of the ulcer base. A localized spqt not protruding from the ulcer base or an arterial jet without a protruding plug or rod was not recorded as a visible vessel. An examination of the ulcer crater was accepted a's ade-

1210 WARA GASTROENTEROLOGY Vol. 88, No.5, Part 1 quate when a visible vessel was either identified or could with certainty be excluded. In order not to interfere with the spontaneous course of the bleeding ulcer, no attempt was made to remove an overlying clot when a water jet had failed to remove it. Endoscopic conferences with photodocumentation of the different stigmata were carried out weekly. The aim of these conferences was to ensure the consistency of the interpretation of the findings. The endoscopies were carried out by senior registrars well trained in diagnostic and therapeutic endoscopy as well as in surgery for bleeding ulcer. Statistical Method Stigmata were related to the subsequent course, with major rebleeding as the outcome criterion. The probability of major rebleeding (D+) when the variable was present (T+) was given by the predictive value positive: p(d+ n T+) TP p(d+it+) - ----. p(t+) TP + FP where TP is the number of true positives and FP the number of false positives. The predictive values positive were compared with the prior probability of major rebleeding, that is, the prevalence of major rebleeding as defined in the present study. Alternatively, the probability of major rebleeding in patients with the variable absent (T-) was calculated according to the following equation: p(d+ n T-) FN p(d+it-) = p(t-) = FN + TN' where FN is the number of false negatives and TN is the number of true negatives (1,2). The relative risk, RR, the probability of major rebleeding with the variable present compared with the probability of major rebleeding with the variable absent, ranks the variables in order of predictive importance: p(d+it+) RR =. p(d+it-) The X 2 method with Yates' correction was used at the 5% level to test for significance of differences between proportions. The Mann-Whitney test was used in the comparison involving interval data. In the evaluation of possible false-negative results, estimating the probability of error of the second kind, a formula for unequal sample size was used (3). Results An actively bleeding ulcer was observed in 99 patients, and the ulcer was marked by stigmata of recent hemorrhage in 174 patients. Twenty-three patients (B%) with a median transfusion requirement of B U of blood presented with arterial uncontrolled bleeding on admission requiring immediate treatment. Thus, 250 patients (92%) with stable circula- tion on admission followed a protocol with hemodynamic assessment of further hemorrhage. The hemorrhage turned out to be self-limited in 190 patients (76%), including 31 patients with minor rebleeds, whereas 60 patients (24%) rebled massively after a median observation time of 14 h. Without reaching significance, patients with oozing hemorrhage were more likely to rebleed massively than patients with stigmata of recent hemorrhage, who tended to stop bleeding spontaneously (Table 1). A subdivision of stigmata of recent hemorrhage, however, disclosed that bleeding was most likely to settle spontaneously in patients presenting with older stigmata (p = 0.02), whereas patients with an overlying clot were almost as likely to rebleed massively as patients with oozing hemorrhage. None of 30 ulcers unmarked by stigmata rebled. A visible vessel, identified in 53 patients, was observed simultaneously with other stigmata in 37 patients and as the only stigma in 16 patients. Patients with a visible vessel, mostly appearing as a plug, rebled more frequently than patients in whom a vessel could not be identified (32% vs. 22%, Table 2), but statistical significance was not reached. Most of the patients with a visible vessel stopped bleeding spontaneously (6B%; 95% confidence intervals 54-BO). Subdividing visible vessels according to bleeding status, the risk of rebleeding increased in all subgroups when a vessel was identified as indicated by the predictive values positive (Table 2). Compared with the prior probability of major rebleeding of 0.24, however, there was an increased risk of rebleeding only in patients with oozing hemorrhage or an overlying clot; this value, however, was not significant. In patients with a visible vessel, bare or together with older stigmata, the risk of rebleeding was lower not only compared with the prior probability but also lower than in patients with oozing or overlying clot Table 1. Predictive Values of Stigmata Subdivided According to Bleeding Status Bleeding status Predictive Self-limited Major value hemorrhage rebleeding positive Active bleeding 52 24 0.32 Oozing, no clot 32 14 0.30 Oozing and clot 20 10 0.33 Recent hemorrhage 138 36 0.21 Clot 69 27 0.28 Older stigmata 56 6 0.10 b Visible vessel bare 13 3 0.19 Total 190" 60 RRa 1.53 1.35 1.47 0.66 1.31 0.34 0.77 a The relative risk (RR) ranks the stigmata in order of prognostic importance. b p = 0.02 compared with a prior probability of major rebleeding of 0.24. "Includes 31 patients with minor rebleeds.

May 1985 ENDOSCOPIC PREDICTION OF MAJOR REBLEEDING 1211 Table 2. Predictive Value of a Visible Vessel Subgrouped According to Bleeding Status Predictive Seif,limited Major value hemorrhijge rebleeding positive.flr" Visible vessel present 36 17 0.32 1.47 Oozing 9 5 0.36 1.~6 Clot 9 8 0.47 2.10 Older stigmata 5 1 0.17 0.69 Visible vessel bare 13 3 0.19 0.77 No visible vessel 154 43 0.22 0.68 Oozing 43 13 19 l1b 0.31 1.42 Clot 60 44 b 19 l1b 0.24 1.00 Older stigmata 51 lb 5 0.09" 0.31 Total 190 58 b 60 22" " The relative risk (RR) ranks the subgroups in order of prognostic importance. b Number with inadequate visualization of ulcer base. "p = 0.02 compared with a prior probability of mqjor rebleed of 0.24. Table 3. Predictive Value of a Visible Vessel Subgrouped According to Ulcer Site Visible vessel present Gastric ulcer 7 6 0.46 2.02 Prepyloric ulcer 15 2 0.12 0.48 Duodenal ulcer 13 9 0.41 1.83 Marginal ulcer 1 No visible vessel Gastric ulcer 54 20" 12 7 c 0.18 0.70 Prepyloric ulcer 44 14" 6 2" 0.12 0.44 Duodenal ulcer 47 19" 25 13" 0.35 1.76 Marginal ulcer 9 5" a None of the predictive values achieved statistical significance compared with a prior probability of major rebleed of 0.24. b The relative risk (RR) ranks the stigma.ta in order of prognostic importance. C Number with inadequate visualization of the ulcer base. Table 4. Predictive Values of a Visible Vessel Related to the Adequacy of the Examination of Ulcer Base without a visible vessel. There was no difference in the distribution of stigmata between ulcer sites. The presence pf a visible vessel in a gastric ulcer "Vas associateq with an increased risk of major rebleeding, but without reaching statistical significance (Table 3). Duodenal ulcers were the most likely to rebleed massively (p < 0.05) and prepyloric ulcers were the most likely to stop bleeding spontaneously (p < 0.05). In contrast to g~stric ulcer, a visible vessel had comparatively less b(;jaring on the course of the hemorrhage in the cases pf prepyloric or duodenal ulcer. Because of active bleeding, clot, or poor endoscopic access an adequate examination of the entire ulcer base was not possible in'80 patients (32%). There was no predominance of poor visualization in pa- Self- Prediclimited tive hem or- Major value rhage rebleeding positive" RRb Predictive Self-limited Major value hemorrhage rebleeding positive" RRb Visible vessel identified 36 17 0.32 1.47 Visible vessel not identified Poor visualization c 58 22 0.28 1.23 Good visualization d 96 21 0.18 0.61 " None of the values achieved statistical significance when compared with a prior probability of major rebleed of 0.24. b The relative risk (RR) ranks the variables in order of prognostic importance. C A visible vessel could have been overlooked. d A visible vessel was with certainty excluded. tients with major rebleeding or duodenal ulcer compared with self-limited hemorrhage or gastric ulcer (37% vs. 31% or 34% vs. 34%, respectively, Table 3). In patients in whom the absence of a visible vessel was established, the risk of rebleeding was decreased compared with the prior probability of major rebleeding (Table 4). In patients with poor visualization the risk of rebleeding was increased but without reaching the same probability as in patients in whom a visible vessel was identified. There was one chance in three of overlooking a visible vessel in the relatively few patients with poor visualization due to active bleeding (f3 = 0.34). For other subgroups of stigmata with poor visualization, includiilg overlying clot, which is the most frequent cause of inadequate examination of the ulcer base, the probability of failing to identify a visible vessel was negligible (f3 :::; 0.05). The median time interval from admission to endoscopy was 2 h for all subgroups of stigmata. Of 190 patients with self-limited hemorrhage and a median transfusion requirement of 2 U of blood, 11 patients (6%) died of an underlying incurable disease. Of the 60 patients with major rebleeding and a median transfusion requirement of 9 U of blood, there were 11 deaths (18%). A comparison of the transfusion requirement and mortality revealed no differences between subgroups of stigmata. Discussion Stigmata of recent hemorrhage were originally used to establish whether a lesion seen at endoscopy actually was the bleeding source (4,5). Later, stigmata were reported to be of prognostic value, predicting a high mortality in patients actively bleeding at endoscopy (6,7). Stigmata have also been claimed to be superior to any other factor in predicting further hemorrhage as well as the need for emergency Sur-

1212 WARA GASTROENTEROLOGY Vol. 88, No.5. Part 1 gery (8). Others, however, have questioned the prognostic importance of stigmata (9). In keeping with other researchers (8,10,11)' we observed that ulcers unmarked by stigmata are unlikely to rebleed. These ulcers were therefore not considered in the evaluation of the prognostic importance of endoscopic stigmata. However, even if focusing only on stigmatized ulcers, most hemorrhages settle spontaneously. The first reports on stigmata did not discuss the relative importance of subgroups of stigmata. Recently, most attention has been paid to the endoscopic appearance of a visible vessel, because a life-threatening ulcer hemorrhage is known to originate from an eroded artery, rather than to nonarterial ulcer blf3eding which is likely to stop spontaneously (12). In patients requiring emergency operation or in postmortem studies the rate of exposed arteries ranges from ~50% (13,14) to 85% (15-17). Furthermore, major hemorrhage is reported to occur predominantly from arteries with an internal diameter >1 mm (17). In contrast, the rate of visible vessels observed at endoscopy is most often reported to be low, ranging from 8% to 48% (10,11,18,19,21). Different timing of endoscopy has undoubtedly contributed to diversity in occurrence of various stigmata in published series. The varying results also reflect interobserver variation and different definitions of stigmata, using different terms and combinations of subgroups of stigmata (8-11,21-23). The visible vessel was first recorded in the presence of a spurting arterial jet or as an obvious projection (18). In some of the subsequent reports the definition of the visible vessel is not clearly presented (19-21). Others have defined the visible vessel in the absence of an arterial jet or p~eudoaneurysm as an elevated raised red or blue spot protruding from the ulcer base (11,22) or simply as a protruding vessel (10). MacLeod et al. (23) included black spots in addition to red and blue spots in the description of the visible vessel, claiming this to be similar to the definition of a visible vessel used by others (11,22). They did not, however, report whether these spots protruded from the ulcer base. A standardization of observational criteria has not yet been proposed but is much needed. The results in published series differ with regard to the prognostic importance of a visible vessel. In a much quoted retrospective study the visible vessel was first brought to notice as a reliable endos(::opic predictor of major hemorrhage (18). In 50% of the patients requiring emergency surgery, however, a visible vessel was not identified at preoperative endoscopy. Uncovering an essentially higher proportion of visible vessels, others observed that only 50% of the visible vessels rebled (11). Although Griffiths et al. (18) claim that the presence of a visible vessel was the crucial finding, Storey et al. (11) concluded that it was the absence of a visible vessel that was important, as other stigmata Without a visible vessel were not likely to hemorrhage further. In controlled studies designed to explore the efficacy of endoscopic methods in controlling hemorrhage, the rate of reble,eding from a small control group of visible vessels has been reported to be as high as 81% but only 50% of the patients required emergency surgery (20). A similar rate of recurrent bleeding has been reported by others (11,22,24). In the series of Vall on et al. (10), 50% of the visible vessels rebled but less than one-third required emergency surgery; this number compares favorably with thf3 rebleeding rate of visible vessels in the present study. In contrast, no reb lee ding was observed from visible vessels as defined by MacLeod et al. (23). Subdividing stigmata according to bleeding status, hemorrhage is reported to recur at low rates when oozing or overlying clots were observed at initial endoscopy (21), particularly in the absence of a visible vessel, but this observation was based on a limited number of patients (11,22,24). The rebleeding rate observed in patients with central spots in the ulcer base (10) compares favorably with the rebleeding rate in patients with stigmata defined as older stigmata in the present study. In patients with nonbleeding visible vessels rebleeding rates of ~ 50% have been reported (10,22); however, differentiation between subgroups of stigmata of recent hemorrhage was not made. The prevalence of the visible vessel in the present study is comparable to most reports (10,18-21,23) but essentially lower than in other studies (11, 22,24). The results in the present study differ from other reports in several important respects, bht primarily in the prognostic importance attached to the visible vessel. The distinction between self-limited hemorrhage and major rebleeding in need of emergency surgery was based on a hemodynamic evaluation as outlined by Andersen (17). The proportion of major rebleeds, the transfusion requirement, and the outcome in patients with self-limited hemorrhage imply a proper distinction (25-27). Rebleeds not fulfilling the hemodynamic criteria of major bleeding and thus not in need of emergency surgery were classified as self-limited hemorrhage. In evaluating the importance of stigmata in the prediction of major rebleeding, patients with minor rebleeds were therefore pooled with patients with no further hemorrhage after admission. The relatively few patients with uncontrolled arterial hemorrhage on admission had emergency hemostasis and were excluded from the protocol, basically

May 1985 ENDOSCOPIC PREDICTION OF MAJOR!{EBLEEDING 1213 because arterial hemorrhage is likely to recur (10,21,23,24,28) and increase the risk of fatal outcome (15,29). A median transfusion requirement of 8 U of blood in patients with arterial bleeding at endoscopy and the fact that all except 1 patient presented with clinical evidence of shock on admission or were transferred from another department after hemodynamic assessment of major hemorrhage reflects the severity of the hemorrhage. It strongly indicates that the major challenge in patients with arterial bleeding at endoscopy is not to predict rebleeding, but to make a decision for treatment based on circulatory response. In contrast, prediction of major rebleeding is of vital importance in the many patients with stable circulation on admission. Relatively few hospitals offer the possibility of endoscopic hemostasis. Most rely on surgery, the currently accepted method in arresting hemorrhage but one which is associated with a significant morbidity and mortality (30,31). The stringent selection for surgery which is therefore imperative emphasizes the importance of predicting major rebleeding reliably. Poor visualization of the ulcer base is a matter of concern. Visibility rates not essentially different from ours have been reported (19,22). The high rate of lesions inaccessible to laser treatment adds to the common problems of adequate visualization (22,23). Timing of endoscopy appears to be a major determinant. The visibility rate in the present study made it necessary to consider the probability of failing to identify a visible vessel in inadequately examined ulcers. The calculation of predictive values and probability of type II-error suggest that it is unlikely that we have failed to identify an essential number of visible vessels despite poor visibility, at least not if it is a potent predictor of major rebleeding as claimed by others (10,11,20,22,24). In conclusion, stigmata or subgroups of stigmata did not emerge as a reliable predictor of major rebleeding. Despite lack of statistical significance, however, there was a trend that bleeding status and ulcer site influence the relative importance of a visible vessel. Hemorrhage was more likely to recur in patients with a visible vessel and oozing or overlying clots or with duodenal or gastric ulcer, whereas a visible vessel bare or identified in patients with older stigmata or prepyloric ulcer rarely rebled. The evaluation of the prognostic importance of stigmata is in its infancy. More prospective data using standardized observational criteria and a hemodynamic distinction between self-limited hemorrhage and major rebleeding are strongly needed. A preliminary report on an endoscopic ultrasound device appears to have promising potential in detecting arterial ulcer hemorrhage (32). For the time being, however, the confined predictive power of stigmata implies that they cannot be recommended in selecting patients for emergency surgery. Whether it is wise to use stigmata in the stratification in controlled trials exploring the efficacy of endoscopic methods in the management of ulcer hemorrhage should also be questioned. References 1. McNeil BJ, Keeler E, Adelstein Sj. Primer on certain elements of medical decision making. N Engl j Med 1975;293:211-5. 2. Wulff HR. Rational diagnosis and treatment. Oxford, London, Edinburgh, Melbourne: Blackwell Scientific, 1976:78-100. 3. Feinstein AR. Clinical biostatistics. Clin Pharmacol Ther 1975;18:491-505. 4. Cotton PB, Rosenberg MT, Waldram RPL, Axon ATR. Early endoscopy of oesophagus, stomach and duodenal bulb in patients with haematemesis and melaena. Br Med j 1973; 2:505-9. 5. Forrest jah, Finlayson NDC, Shearman Dje. Endoscopy in gastrointestinal bleeding. Lancet 1974;ii:394-7. 6. Nudel J, Guarena J, Milman PJ, et al. Endoscopic diagnosis of active bleeding: a prognostic sign in upper gastrointestinal hemorrhage. Gastrointest Endosc 1977;23:237(A). 7. Gilbert DA, Silverstein FE, Tedesco FJ, Buenger NK, Persing J, et al. The national ASGE survey on upper gastrointestinal bleeding. Part III. Endoscopy in upper gastrointestinal bleeding. Gastrointest Endosc 1981;27:94-102. 8. Foster DN, Miloszewski KJA, Losowsky MS. Stigmata of recent haemorrhage in diagnosis and prognosis of upper gastrointestinal bleeding. Br Med j 1978;1:1173-7. 9. MacLeod la, Mills PRo Factors identifying the probability of further haemorrhage after acute upper gastrointestinal haemorrhage. Br j Surg 1982;69:256-8. 10. Vall on AG, Cotton PB, Laurence BH, Miro JRA, Oses jcs. Randomized trial of endoscopic argon laser photocoagulation in bleeding peptic ulcer. Gut 1981;22:228-33. 11. 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. 12. Osborn GR. The pathology of gastric arteries with special reference to fatal haemorrhage from peptic ulcer. Br J Surg 1954;41 :585-94. 13. Djorup F. The source of bleeding in acute gastric haemorrhage. Acta Chir Scand 1965;(Suppl 343):102-12. 14. Chalmers TC, Zamcheck N, Curtins GW, White FW. Fatal gastrointestinal hemorrhage: clinicopathologic correlations in 101 patients. Am J Clin Pathol 1952;22:634-5. 15. Jones FA. Haematemesis and melaena with special reference to bleeding peptic ulcer. Br Med j 1947;2:441-6. 16. Pedersen J. Lethality rate of hematemesis and melena treated non-operatively (Meulengracht's regimen) and criteria for surgical intervention in bleeding peptic ulcer. Gastroenterology 1949;12:597-616. 17. Andersen D. The use of measurement of central venous pressure in the selection of patients with massive gastroduodenal haemorrhage for emergency operation. Scand J GastroenteroI1970;5:25-32. 18. Griffiths WJ, Neumann DA, Welsh jd. The visible vessel as indicator of uncontrolled or recurrent gastrointestinal hemorrhage. N Engl j Med 1979;300:1173-7. 19. Fleischer D. Etiology and prevalence of severe persistent upper gastrointestinal bleeding. Gastroenterology 1983;84: 538-43.

1214 WARA GASTROENTEROLOGY Vol. 88, No.5, Part 1 20. Papp JP. Electrocoagulation in upper gastrointestinal bleeding. Dig Dis Sci 1981;26:41-3. 21. Rutgeerts p, Vantrappen G, Broeckaert L, et al. Controlled trial of YAG laser treatment of upper digestive hemorrhage. Gastroenterology 1982;83:410-6. 22. Swain CP, Bown SG, Storey DW, Kirkham JS, Northfield TC, Salmon PRo Controlled trial of argon laser photocoagulation in bleeding peptic ulcers. Lancet 1981;ii:1313-6. 23. MacLeod la, Mills PR, MacKenzie JF, Russel RI, Carter DC. Neodymium Y AG laser photocoagulation for major haemorrhage from peptic ulcers and single vessels: a single blind controlled study. Br Med J 1983;286:345-8. 24. Swain CP, Bown SG, Salmon PR, Kirkham JS, Northfield TC. Controlled trial of Nd YAG laser photocoagulation in bleeding peptic ulcers. Gastrointest Endosc 1984;30:137(A). 25. Wara P, Stodkilde H. Bleeding pattern prior to admission as guideline for emergency endoscopy. Scand J Gastroenterol (in press). 26. Wara P, Host V, St0dkilde H. Clinical factors predisposing to major ulcer hemorrhage. A logistic regression analysis. Acta Chir Scand (in press). 27. Wara P. Endoscopic electrocoagulation of major ulcer hemorrhage. Acta Chir Scand (in press). 28. Andersen DA. The use of central venous pressure measurement in the diagnosis of major arterial bleeding in gastroduodenal haemorrhage. Scand J Gastroenterol 1969;4:529-36. 29. Jones PF, Johnston SJ, McEwan AB, Kyle J, Needham CD. Further hemorrhage after admission to hospital for gastrointestinal haemorrhage. Br Med J 1973;3:660-4. 30. Dronfield MW, Atkinson M, Langman MJS. Effect of different operation policies on mortality from bleeding peptic ulcer. Lancet 1979;i:1126-7. 31. Wara P, Berg V, Amdrup E. Factors influencing mortality in bleeding peptic ulcer. Review of 7 years' experience preceding therapeutic endoscopy. Acta Chir Scand 1983;149:775-85. 32. Beckly DE, Casebow MP. Preliminary clinical experience with an endoscopic Doppler ultrasound device in Gl haemorrhage. Gut 1983;24:968(A).