Nature of the Bleeding Vessel in Recurrently Bleeding Gastric Ulcers

Transcription

1 GASTROETEROLOG 1986;90: ature of the Bleeding Vessel in Recurrently Bleeding Gastric Ulcers C. PAUL SWAI, DAVID W. STORE, STEPHE G. BOW, JEA HEATH, TIMOTH. MILLS, PAUL R. SALMO, TIMOTH C. ORTHFIELD, J. SQUIRE KIRKHAM, and JERR P. O'SULLIVA orman Tanner Gastroenterology Unit and Department of Histopathology, St. James' Hospital and Department of Gastroenterology, University College Hospital, London, United Kingdom An unselected consecutive series of 826 patients admitted for acute upper gastrointestinal bleeding underwent urgent endoscopy. Peptic ulcers were found in 402 (49%). Of the 329 ulcer craters that could be fully examined, visible vessels were identified in 156 (47%), other stigmata of recent hemorrhage in 66, and no stigmata of recent hemorrhage in 107. One hundred twenty-nine patients with stigmata of recent hemorrhage (93 of whom had visible vessels) randomly allocated to no endoscopic treatment were observed for evidence of further bleeding. Fifty-four of the 93 patients (58%) with visible vessels rebled, compared with 2 of 36 [6%) with other stigmata of recent hemorrhage. o patient without stigmata of recent hemorrhage rebled. Twenty-seven patients in whom a visible vessel in a gastric ulcer was identified at endoscopy underwent urgent partial gastrectomy because of recurrent bleeding. The vessel identified at endoscopy was found in 26 of 27 resection specimens (96%). The arterial vessel wall protruded above the surface of the ulcer crater in 10 specimens, and clot in continuity with a breach in the vessel wall protruded in a further 10 specimens. Postoperative angiography, when technically possible, showed that the breached artery ran across the base of the ulcer in all of these specimens. Pathological changes were common in the bleeding artery and included arteritis in 24 of 29 (83%) eroded arteries found in these specimens, with aneurysmal dilatation in 14 of 27 (52%) bleeding points that could be fully examined. The ulcer had penetrated to serosa in 13 specimens (45%). The bleeding artery Received August 17, Accepted September 4, Address requests for reprints to: Dr. C. Paul Swain, St. James' Hospital, Sarsfeld Road, London S.W. 12, United Kingdom. This study was supported by a grant from the Department of Health and Social Security by the American Gastroenterological Association /86/$3.50 had a mean external diameter of 0.7 mm with a range of mm. This study provides new information about the nature of the bleeding vessel in gastric ulcers, and some of this information is relevant in planning studies of endoscopic therapy for bleeding peptic ulcers. It validates the endoscopic identificatio'n of a visible vessel, and confirms that such identification has a high predictive value for the development of recurrent hemorrhage. In 1829 Cruveilhier (1) described the appearance of a vessel protruding from the floor of an ulcer in patients dying of gastrointestinal bleeding. In the 1930s endoscopists in Germany considered the prognostic significance of a vessel visible in the base of an ulcer that had recently bled (2), and also described stigmata of recent hemorrhage (SRH), i.e., blood adherent to an ulcer that had recently bled (3). Recently, the endoscopic observation of a visible vessel (4, 5) and of other SRH (6) has been stressed as a valuable predictive indicator of the risk of further bleeding and death. More observations concerning the nature of the bleeding vessel in massively bleeding ulcers would be of value to clinicians interested in endoscopic control of gastrointestinal hemorrhage, and might validate the endoscopic diagnosis of the visible vessel. White scar tissue might reflect laser light or prevent effective thermal contraction of the bleeding vessel. Vessels with diameters >1 mm are difficult to seal (7-9) using the currently available thermal methods (such as lasers or electrocoagulation). Measurements of stomach wall thickness are relevant to the safety of endoscopic therapy. The aim of this study was to characterize the Abbreviation used in this paper: SRH, stigmata of recent hemorrhage.

2 596 SWAI ET AL. GASTROETEROLOG Vol. 90. o.3 endoscopically observed visible vessel, using histologic and angiographic observations in gastric resection specimens, in order to provide quantitative information on the size of the bleeding vessel, on pathological changes in the bleeding vessel, and on the floor of gastric ulcers with continued or recurrent bleeding. Materials and Methods Clinical Material The clinical material for this study was derived from 826 consecutive patients admitted because of acute upper gastrointestinal bleeding during a 4-yr period from 1979 to All patients underwent urgent endoscopy. If an ulcer was found, the patients were prospectively grouped into those with a visible vessel, those with minor SRH, and those without SRH. The ulcer crater was gently washed with water to remove overlying fresh blood. mucus, loose clots, or debris in order to obtain a view of the entire ulcer crater. The 402 peptic ulcers seen at urgent endoscopy in this series included 6 esophageal ulcers, 181 gastric ulcers. 198 duodenal ulcers, and 17 stomal (jejunal) ulcers. In the absence of spurting arterial bleeding or a pulsating pseudoaneurysm indicating the site of a vessel, a visible vessel at endoscopy was defined as an elevated red or black spot that protruded from the ulcer crater, was resistant to washing, and was often associated with the freshest clot in the ulcer crater. When no vessel was visible. the ulcer was regarded as having SRH if oozing was seen and if fresh blood or nonelevated red or black spots were seen in the crater. Rebleeding while in the hospital, as judged by a clinician independent of the endoscopist. was taken as an indication for urgent surgery. Evidence of rebleeding was regarded as definitive if fresh hematemesis occurred. but as suggestive if there was fresh melena. a sudden rise in pulse rate. or a decrease in blood pressure or central venous pressure. In cases of suggestive evidence. rebleeding was confirmed by repeat endoscopy. This study was carried out during the course of two controlled trials of laser photocoagulation (10.11) for hemorrhage from bleeding peptic ulcer: patients who had ulcers with visible vessels or other SRH were prospectively and separately randomized to laser treatment or no endoscopic treatment, and patients without SRH did not receive endoscopic treatment. There was a 3-mo gap between trials of argon and neodymium:yttrium aluminum garnet (d:ag) laser photocoagulation while new d:ag lasers were installed. During this period, patients consecutively included in the present study entered the no endoscopic treatment group. Pathological Material The pathological material for this study consisted of the 27 gastrectomy specimens derived from this consecutive series of patients with bleeding gastric ulcers who required urgent gastrectomy after continued or further bleeding subsequent to the endoscopic finding of a visible vessel in the floor of the gastric ulcer. The vessels in gastric ulcers in these 27 patients were diagnosed at endoscopy but were not treated endoscopically. Because of the nature of urgent surgery for bleeding duodenal ulcer, where oversewing of the bleeding point with vagotomy or gastrectomy was performed. no duodenal specimens containing the ulcer or vessel seen at endoscopy were available for study. Preparation of the Specimen Angiography of resected specimens. The fresh specimen was examined to identify the closest large artery feeding the serosal surface near the bleeding ulcer. The largest appropriate artery was cannulated. the blood was washed out with saline. and the artery was then injected with a mixture of barium sulfate suspended in hot gelatin. The injection pressure was maintained until barium and gelatin spurted from the orifice of the eroded vessel in the ulcer, or until the arteries on the serosal side under the ulcer were filled with barium. Barium in the ulcer was wiped away and the gelatin was allowed to set. Macroradiographs were taken using mammography equipment and the bleeding point was marked with a pointed 23 gauge-, 0.6-mm-diameter needle for comparative measurements of vessel size. Oriented sections were cut along the axis of the vessel allowing lateral radiographs to be taken, so that the vessel course could be followed in three dimensions. Histologic Methods The stomach was pinned out. stretched flat. and placed in 10% formol saline as soon as the radiology was completed. The entire ulcer was sectioned in the axis of the vessel and the blocks were marked so that a three-dimensional picture of the vessel path could be reconstructed. After routine processing in paraffin wax. 5-lLm step-serial sections through the entire thickness of each block were cut and stained with hematoxylin and eosin and with elastic Van Gieson's stain. In selected cases. sections were stained by the M.S.B. method for fibrin. Histologic Measurements Using a camera lucida attachment to a conventional microscope, the histologic section was projected onto a high-resolution data tablet (model MOP/AM01, Kontron Analytical, Everett. Mass.). Using the special mechanical pen that is part of the system and that interacts electronically with the data tablet. the lengths of each parameter were determined by tracing across the feature in question. This system allowed rapid and accurate determination of lengths. and reproducibility of the measurements was ensured by repeating each estimation 10 times. The magnification factor of the projection was determined using a slide calibrated in divisions of m. Because the artery was sometimes aneurysmally dilated at the bleeding point. it was followed and measured at 1 mm from the bleeding point to determine its caliber. As an

3 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 597 artery may be cut at an angle, the minimum transverse diameter was taken as the true reading, and ellipsoidal cuts through a vessel were avoided. Some reservations should be expressed about the nature of absolute measurements of highly elastic structures. The diameter of arteries varies in systole and diastole, and may contract upon surgical removal and histologic fixation. The stomach wall thickness varies depending on the degree of distention of the stomach, which normally and during endoscopy varies widely in relation to the stomach contents or the degree of insufflation with air. onetheless, measurements at histology are precise and allow valid quantitation and accurate comparison, provided caution is exercised in extrapolating from these measurements to values in vivo. Statistical Methods The significance of observed differences was assessed by.i analysis (12) with ates' continuity correction (13), by Fisher's exact test for small numbers (14), and by Student's t-test to test differences between means (15). Results Clinical Observations The incidence and prognosis of the visible vessel and other SRH, as observed endoscopically in patients with peptic ulcers admitted to the hospital because of gastrointestinal bleeding, were assessed prospectively in the following clinical material (Figure 1). Peptic ulcers were found in 402 of the 826 patients who underwent endoscopy (49%). Full endoscopic examination of the crater was not possible in 73 of 402 (18%) of these ulcers, visible vessels were identified in 156 (39%), other SRH were observed in 66 (16%), and no SRH was observed in 107 (27%). The prognostic significance of these endoscopic findings was assessed prospectively in 129 patients with SRH randomly allocated to no endoscopic treatment and observed for evidence of continued or recurrent bleeding. Fifty-four of 93 patients (58%) with visible vessels rebled compared with 2 of 36 patients (6%) with other SRH (p < 0.001). o patients without SRH rebled. Urgent surgery was required in 48 of 93 patients with visible vessels compared with 1 of 36 patients with other SRH (p < 0.005). Sixteen of 93 patients with visible vessels died compared with 1 of 36 patients with other SRH (p < 0.02). Thus occurrences of further bleeding, urgent surgery, and mortality were almost entirely restricted to patients with ulcers having visible vessels. o patients without SRH had further bleeding, required urgent operation, or died. Qualitative Histologic Observations Qualitative histologic observations on the nature of the bleeding vessel and on the ulcer crater in 27 gastrectomy specimens are presented in Table 1. A structure recognized histologically as an arterial vessel protruded above the surface of the ulcer crater (Figure 2) in 10 of 27 specimens (7 of 10 protruding vessels had an adherent clot). In another 10 specimens only a clot protruded above the surface, either as a plug in continuity with the vessel (Figure 3) or forming a pseudoaneurysmal roof on the vessel (Figure 4). In the remaining 7 specimens the vessel was flush with the surface of the ulcer (Figure 5). Unequivocal aneurysmal dilatation of the vessel at the bleeding point was common, occurring in half (14 of 27) of the bleeding vessels studied (Figure 6). Sometimes a remarkable relative increase in the diameter of the bleeding vessel was observed. The dilatation was eccentric in all cases, with the swelling on the eroded luminal side of the artery. Although occasional reports of aneurysmal dilatation of an artery in a bleeding gastric ulcer have appeared (16,17), this is the first report to establish the great frequency of aneurysmal dilatation as a pathological finding in bleeding gastric ulcers. Pathological change was common in the exposed artery. An intense arteritis with polymorphonuclear cell infiltrate and fibrinoid necrosis was observed in most eroded arteries (24 of 29). Occasionally, it was difficult to determine where fibrinoid necrosis of the vessel wall merged into thrombus. Recanalized thrombus (Figure 7) was observed in 7 of 29 vessels, indicating that the artery had bled, stopped bleeding as thrombus formed, and had then rebled. Although atheroma was never observed in these bleeding arteries, a proliferative change (designated as loose intimal thickening) was sometimes seen in the bleeding artery (3 of 28 vessels). This change proved fairly frequent when it was looked for elsewhere in the section of the ulcer (19 of 29) (Figure 8) and was also found in sections away from the ulcer (8 of 21), though less frequently. In the majority of patients in this series, bleeding was caused by the erosion of a single artery at the floor of an ulcer-small vessels of <0.1 mm in diameter were also breached in 10 of 29 bleeding ulcers. In 1 patient, a single artery was eroded at two points (4 mm apart) (Figure 9). In another patient, three ulcers were identified and an eroded artery was found in each. It is of interest that the accompanying vein was also breached by the ulcerating process in 4 cases. Figure 10 shows an unusual case (not from this series) where a gastric ulcer was shown to be bleeding from an eroded vein with an intact accompanying artery.

4 826 patients with UGI bleeding i / 402 total of patients with ulcers 424 patients with lesions other than peptic ulcers 329 ulcers fully examined 73 ulcer craters not fully examined 222 ulcers have SRH (156 visible vessel, 66 other SRH) ulcers with SRH are randomly allocated to no endoscopic treatment and are observed for evidence of further bleeding ulcers with SRH have a visible vessel 107 ulcers have no SRH (no patient without SRH rebled) - 93 patients with SRH randomly enter active arm of therapeutic endoscopic laser trial - 36 patients have other SRH (2 of 36 rebled) 54 ulcers with a visible vessel have further bleeding (27 GU, 27 DU) 39 ulcers with a visible vessel have no further bleeding 27 patients undergo urgent gastrectomy for recur- 21 patients undergo urgent oversewing with varent bleeding from a gastric ulcer gotomy or gastrectomy for recurrent bleeding / from a duodenal ulcer 27 gastrectomy specimew from this consecutive series enter this study - 6 are managed conservatively, i.e, nonoperatively Figure 1. A flow diagram to show how the gastrectomy material from patients with continued or recurrent hemorrhage from gastric ulcers was derived from a consecutive series of 826 patients with upper gastrointestinal bleeding. The diagram indicates the incidence of rebleeding, visible vessels, stigmata of recent hemorrhage, and gastric (CU) and duodenal (DU) ulceration in the patients with peptic ulceration from this series. '" <0 CD en ~ Z t%] >-l > r C'l > en ~ Z >-l t%] :::tl o r o C'l ><: <: 1?- <0 '? Z ~ w

5 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 599 Table 1. Histologic Observations on the Bleeding Vessel and Ulcer Crater Floor in 27 Gastrectomy Specimens o a 7b 8 9a 9b 9c Total Mean 0.66 b b b Stomach wall thickness (mm) Vessel Artery above Clot diam- ulcer above eter surface ulcer (mm) surface Arter- His Aneurysm Recanalized thrombus 10/27 17/27 14/27 24/29 7/29 Accompanying Small LIT Serosal vein vessels at artery breached breached BP 10/29 4/29 LIT Muscuelse- LIT laris where distal propria in sec- to ultion penetrated cer Scar tissue 10/29 3/28 19/29 8/21 13/29 17/29 a Indicates instances where reliable observations could not be made. b Indicates that no bleeding vessel was found in this specimen. C Indicates that no sections were available for examination. In patient o.7. two bleeding points were found in one ulcer. In patient o. 9, three ulcers were found, each with a bleeding point. LIT, loose intimal thickening; BP, bleeding point;, no;, yes. Arterial diameter was measured 1 mm from the bleeding point. The thickness of the stomach wall was measured 1 em from the ulcer edge The bleeding artery was usually eroded in the lowest layer of tissue penetrated by the ulcer; occasionally (3 of 29 cases) a submucosal artery was found bleeding close to the edge of an ulcer that had penetrated the muscularis propria. More than half of the bleeding ulcers were of acute type with penetration only into the submucosa (16 of 29). The majority of these acute ulcers had no scar tissue (12 of 16). Scar tissue occurred significantly more frequently in ulcers that penetrated the muscularis propria (13 of 13) than in those that did not (4 of 16) (p < 0.001). An example of an aneurysm ally dilated serosal artery, looping up to protrude above the surface of a chronic, heavily scarred ulcer that has penetrated through the muscularis propria, is shown in Figure 11. It proved possible to measure the diameter of the bleeding artery 1 mm away from the bleeding point in 25 of these specimens (mean arterial diameter, 0.66 mm; range, mm; standard devia- tion, 0.44 mm). In 5 of 25 cases, the diameter of these arteries was> 1 mm. The vessel wall thickness could be measured in 26 arteries (mean diameter, 0.19 mm; range, mm; standard deviation, 0.15 mm). The thickness of the whole stomach wall and the muscularis propria were measured 1 cm from the ulcer. The thickness of the whole stomach wall was measured in 26 specimens (mean, 4.73 mm; range, mm; standard deviation, 1.62 mm). The mean thickness of the muscularis propria was also i Quantitative Histologic Observations

6 600 SWAI ET AL. GASTROETEROLOG Vol. 90, o.3 Figure 2. This artery protrudes above the surface of the ulcer crater, suggesting that the vessel wall may sometimes be more resistant to acid digestion than other structures in the floor of the ulcer (Reference 10). measured in 26 specimens (mean, 2.23 mm; range, mm; standard deviation, 0.99 mm). Angiographic Observations and Measurements Angiographic observations were made on the 13 specimens in which barium and gelatin could be successfully injected into the arterial system supplying the bleeding point. Using this system it was possible to demonstrate that the artery crossed the ulcer floor in all. Injection of two large-diameter bleeding arteries emptied all the barium into the ulcer without crossing to the other side, but with a probe it was easy to demon- strate that the artery continued across the floor of the ulcer, i.e., that these were not "end arteries." Lateral radiographic views of arterially injected sections cut through the ulcer (Figure 12) and radiographs in the anteroposterior view (Figure 13) allowed a three-dimensional picture of the path of the bleeding vessel to be correlated with the histologic picture. The artery was observed to loop upward in a vertical plane toward the bleeding point in the floor of the ulcer in 7 of 13 cases. There was kinking in a lateral plane at the bleeding point (e.g., in Figure 13) in 4 of 13 cases. Small vessels were shown to be breached in 3 of 13 cases (e.g., in Figure 11). Aneurysmal dilatation was demonstrated in 5 of 13 cases. A 0.6-mm needle was placed with its tip on the bleeding point to: allow measurement of vessel diameter. The mean internal diameter of 0.38 mm correlated well with the mean internal diameter of 0.3 mm measured at histology. This scale comparison also allowed measurements of the mean distance between branch points from the main injected trunk (usually the left gastric artery) to the bleeding point (3.5 cm), and the mean number of branch points between the main trunk and the bleeding point (3). Figure 3. A plug of thrombus in continuity with the wall of the artery protrudes above the floor of the ulcer. The layers of the protruding thrombus give an appearance mimicking the presence of a lumen. In this case the artery wall was eroded flush with the ulcer crater floor (Reference 14). Discussion Analysis of this clini1::al material confirms that the identification of a visible vessel at urgent endoscopy carries important adverse prognostic significance. The occurrence of further bleeding and mortality are almost entirely restricted to patients with ulcers having visible vessels, but only half of these patients have further bleeding. Exposed arteries that have recently bled have been described by pathologists since Cruveilhier (1,18,19) and are reported as a common finding in several series (17,20-22). For example, in the retrospective pathological study by Chalmers et al. (23), of 101 patients dying of upper gastrointestinal tract bleeding with grossly apparent blood visible in the gastro-

7 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 601 Figure 4. The artery loops up toward the floor of the ulcer and turns sharply at the bleeding point to run down into the submucosa. Aneurysmal dilatation of the artery with arteritis and fibrinoid necrosis can be seen at the bleeding point. A "false" roof of clot protrudes above the rest of the ulcer floor (Reference 2). Figure 5. This artery runs just below the floor of the ulcer without looping. The arterial wall has been eroded over a considerable distance (6 mm) (Reference 5).

8 602 SWAI ET AL. GASTROETEROLOG Vol. 90, o.3.,., -,,, Figure 6. There is "massive," 3.B-mm, true aneurysmal dilatation of this 0.7-mm-diameter submucosal artery at the bleeding point (Reference 22). intestinal tract at autopsy, more than half (26 of 50) of patients with benign ulcers were recorded as having a blood vessel exposed in the ulcer. The observations in this prospective study go further in showing that an eroded artery can almost always be found at histology, as well as at endoscopy, in gastric ulcers that have further bleeding and require surgery. The structure protruding from the ulcer, identified at endoscopy as a visible vessel, was not always the Figure 7. There is recanalized thrombus in the artery (Reference 6).

9 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 603 Figure 8. The three sections through this artery show loose intimal thickening. The margins of this change are indicated in the right-hand section by curved arrows. The branch at the top left, marked with a straight arrow, does not show this feature, indicating the focal nature of this change. The intima should be, at most, a few cells thick as in this branch. Figure 9. This submucosal artery crosses immediately below the floor of this ulcer without looping upward. Two pseudoaneurysmal "blowouts" are separated by a distance of 4 mm. Remnants of the arterial wall protrude above the surface of the ulcer in the aneurysm on the right. Clot alone protrudes above the surface of the ulcer in the aneurysm on the left (Reference 7).

10 604 SWAI ET AL. GASTROETEROLOG Vol. 90. o.3 Figure 10. A large vein is eroded in the floor of this ulcer. The intact accompanying artery can be seen by its side. vessel wall. Sometimes it was a clot in continuity with the eroded vessel wall that formed a plug or pseudoaneurysmal roof on the vessel. The endoscopic concept of the visible vessel is thus something of a simplification when analyzed at histology. The vessel wall is in fact almost transparent, and hence difficult to distinguish from other pale structures in the ulcer floor; the dark red raised appearance that the endoscopist recognizes is either blood within the vessel or, more commonly, is a clot in continuity with the eroded vessel, plugging a hole or roofing an aneurysm. Aneurysmal dilatation was found at the bleeding point with striking frequency in this series [14 of 27 cases, (52%)]. It is probable that aneurysm formation due to the effects of the ulceration is even more common than is implied by these figures, as the aneurysmal roof may be blown off during a bleed and as histologic preparation may not always demonstrate aneurysm formation clearly. Pathological change was common in the bleeding artery. In the majority of cases (24 of 29) an arteritis was seen in the wall of the artery adjacent to the floor of the ulcer. The obliterative endarteritis of small vessels in specimens from patients with bleeding ulcers, occasionally described in older texts (24-26), was observed only once in a single section in this series. In chronic and some acute ulcers there may be time for reactive changes to take place in the vessels involved. Recanalized thrombus was occasionally observed (7 of 29 cases), suggesting that the artery had bled, had stopped bleeding as thrombus formed, and had then rebled. A proliferative change, designated as loose intimal thickening, was sometimes found in the eroded artery and was common in arteries elsewhere in the section of the ulcer, but was less common in arteries in sections distal to the ulcer. This finding did not correlate with the age of the patient or with clinical or electrocardiographic evidence of ischemic heart disease. It has been suggested (26-28) that age-related changes and atheroma in the bleeding vessel might prevent normal vessel contraction and might explain why elderly patients with gastrointestinal bleeding are at increased risk of death. In our series no vessels were observed to have changes that could be interpreted as atheroma. These data support those of Osborn (22). Age-related change appears to be unimportant in the pathogenesis of bleeding gastric ulcers, in keeping with the clinical observation that the

11 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 605 Figure 11. This section shows a serosal artery looping upward to protrude with aneurysmal dilatation above the floor of a chronic, heavily scarred ulcer, which has penetrated through the muscularis propria seen on both sides of the ulcer. incidence of rebleeding following hospital admission is unrelated to age (29). The size of the bleeding artery may be a very important determinant of outcome following a gastrointestinal bleed. Blood flow through an artery varies with the fourth power of the internal luminal diameter of that artery. A 1.2-fold increase in the luminal diameter would double the blood flow. Doubling the luminal diameter would increase the blood flow 16-fold. A calculation of approximate Figure 12. This macroradiograph of a lateral view of a section through a chronic ulcer, in which the arterial system has been injected with a mixture of barium and gelatin that has been allowed to set, shows the large-diameter bleeding serosal artery in the floor of the ulcer marked by the arrow. Small vessels can also be seen to be breached in the granulation tissue at the top right edge of this ulcer (Reference 18).

12 606 SW AI ET AL. GASTROETEROLOG Vol. 90, o.3 Figure 13. This anteroposterior radiograph of a gastric resection specimen after injection with barium and gelatin shows that this artery branches three times before reaching the bleeding point marked with the tip of the 0.6-mm needle. The internal diameter of the bleeding artery can be seen to be 0.6 mm. Kinking in the lateral plane at the bleeding point is also observed (Reference 10). blood loss from an artery of 0.3-mm luminal diameter (the mean of measurements in this study) gave a figure of 67 mllh. If the luminal diameter were 0.6 mm (as in Figure 13), the rate of calculated blood loss would be 1066 mllh if parameters of vessel length, branching, and elasticity were not altered. The blood loss from the erosion of an artery < 0.1 mm in diameter can rarely be sufficient to constitute a threat to life, while normal hemostatic mechanisms can seal an eroded artery >1 mm in diameter only with difficulty because of the high blood flow. These observations carry implications for clinicians interested in endoscopic therapy for gastrointestinal bleeding. The majority of these bleeding arteries were <1 mm in diameter. Lasers and endoscopic electrodes are effective in coagulating normal arteries :51 mm in diameter in animal models (7-9), but become less effective as the arterial diameter increases above 1 mm. Arteries >1 mm in diameter were occasionally encountered in this series as a source of bleeding; these cannot be sealed by conventional available means, and effective endoscopic means to seal such vessels must be developed. The presence of aneurysm may explain why thermal endoscopic methods sometimes precipitate arterial bleeding, and also lends support to the empirically derived method of treating the area around the circumference of an exposed vessel, as this will increase the chance of thermally coagulating the normal vessel without disturbing the aneurysm. It may not be wise to attempt to gauge the size of the bleeding vessel from the luminal side at endoscopy or surgery. The artery was always in the floor of the ulcer, though not necessarily in its center. Oozing from tissue on the mucosal edge of the ulcer is unlikely to be a significant source of major bleeding, and treatment delivered to this area is unlikely to be rewarding. The artery was never severed by the ulcerating process. This vitiates an important normal mechanism of hemostasis-that of retraction contraction of a severed vessel. The artery may bleed from either side of the orifice as there are rich anastomoses between gastric and duodenal arterial arcades. Angiography demonstrated that it was sometimes possible to fill the same bleeding artery by injecting into different serosal arteries. Treatment may be required to seal the artery proximally and distally to the bleeding point. The common tendency of the artery to loop up to the surface (illustrated in Figures 4, 11, and 12) produces a configuration that makes it difficult to treat normal parts of the bleeding artery.

13 March 1986 BLEEDIG VESSEL I GASTRIC ULCERS 607 In this series the high incidence of acute ulcers that are submucosal (19 of 29) and have no scar tissue in their floor (15 of 19) is of special interest, indicating that in a majority of cases scar tissue is not present to prevent effective thermal contraction of the vessel. Bleeding from serosal arteries (Figure 13) in chronic ulcers was not uncommon (10 of 29). Classic studies by Rokitansky (30) have indicated that in bleeding from chronic ulcers in the upper gastrointestinal tract, the bleeding vessel frequently may be serosal (Le., extragastric). Serosal arteries breached by ulcers in this series were generally larger than submucosal arteries (mean serosal arterial diameter, 0.88 mm; range, mm vs. mean submucosal arterial diameter, 0.50 mm; range mm; t = 2.092, P < 0.05). Treating arteries that are technically outside the stomach (the ulcer having perforated all layers of the stomach) may appear dangerous but, paradoxically, is probably safer than treating submucosal arteries as the ulcer is usually stuck down firmly to other structures by a thick mat of scar tissue. Measurement of stomach wall thickness is relevant to the safety of endoscopic therapy. Thickness of the stomach wall varied greatly but was sometimes very thin, one stomach wall being only 2 mm thick. It is probable that more deeply penetrating thermo coagulative methods, such as the d:ag laser and monopolar electrocoagulation, would cause full-thickness tissue damage in such stomachs. These findings demonstrate that the bleeding point in an ulcer can almost invariably be identified both at endoscopy and at histology, and validate the endoscopic recognition of the visible vessel as a useful indicator of high risk of further bleeding. This detailed analysis of the nature of the bleeding vessel puts the phenomenon of the bleeding peptic ulcer on a firmer pathological basis. Appendix Glossary of Endoscopic and Histologic Terms Acute ulcer: an ulcer with no granulation tissue in the crater floor. Aneurysm: a localized eccentric or concentric dilatation of an artery. Atheroma: deposition of fibrofatty plaques in the intima. Arteritis: Infiltration of the vessel wall by inflammatory cells. Chronic ulcer: an ulcer with granulation or scar tissue in the crater floor, which has usually but not always penetrated the muscularis propria. Fibrinoid necrosis: degeneration of the vessel wall showing positive staining for fibrin. Loose intimal thickening: a focal increase in the distance between the internal elastic lamina and the endothelium filled by loose connective tissue. Recanalized thrombus: formation of new endothelial lined channels through a thrombotic occlusion. Ulcer at endoscopy: a break in the mucosa >0.5 mm in diameter. Visible vessel: an elevated red or black spot that protrudes from an ulcer crater, is resistant to washing, and is often associated with the freshest clot in the ulcer crater. References 1. Cruveilhier J. Anatomie pathologique du corps humain. Paris: Balliere, 1829: Schindler R. Gastroscopy. Chicago: The University of Chicago Press, Gutzeit K, Tietge H. Die Gastroskopie. Berlin, Vienna: Urban & Schwarzenberg, 1933: Griffiths WJ, eumann DA, Welsh JD. The visible vessel as an indicator of uncontrolled or recurrent hemorrhage. Engl J Med 1979;300: Storey DW, Bown SG, Swain CP, Salmon PR, Kirkham JS, orthfield TC. Endoscopic prediction of recurrent bleeding in peptic ulcers. Engl J Med 1981;305: Foster D, Miloszewski KJA, Losowsky MS. Stigmata of recent haemorrhage in diagnosis and prognosis of upper gastrointestinal bleeding. Br Med J 1978;1: Gorisch W, Boergen K-P. Heat-induced contraction of blood vessels. Lasers Surg Med 1982;2: Mills T, Swain CP, Dark JM, Morrison S, Salmon PRo The "hot squeeze" bipolar forceps. A more effective endoscopic method for stopping bleeding from large vessels in the gastrointestinal tract. Gastrointest Endosc 1983;29:184-5A. 9. Johnston J, Jensen D, Auth D. Comparison of endoscopic lasers, electrosurgery and the heater probe in coagulation of canine arteries. Gastrointest Endosc 1984;30:154A. 10. Swain CP, Bown SG, Storey DW, orthfield TC, Kirkham JS, Salmon PRo Controlled trial of argon laser photocoagulation in bleeding peptic ulcers. Lancet 1981;ii: Swain CP, Bown SG, Salmon PR, Kirkham JS, orthfield TC. Controlled trial of d AG laser photocoagulation in bleeding peptic ulcers (abstr). Gastroenterology 1983;84: Pearson K. On the criterion that a given system of deviations from the probable in the case of a correlated system of variables is such that it can be reasonably supposed to have arisen from random sampling. Philos Mag 1900;1: ates F. Contingency tables involving small numbers and the x"-test. J R Stat Soc B 1934;1 (Suppl.): Fisher RA. Statistical methods for research workers. 5th ed. Edinburgh: Oliver and Boyd Ltd., "Student" (Gossett WS). The probable error of a mean. Biometrika 1908;6: Tanner C. The diagnosis and management of massive haematemesis. Br J Surg 1964;300: Tixier L, Clave I C. Les grandes hemorragies gastroduodenales. Paris: Masson et Cie, Libraire de l'academie de Medicin, 1933: Murchison C. Two cases of fatal haematemesis from very minute ulcers perforating a small artery in the coats of the stomach. Trans Pathol Soc London 1870;21; Dieulafoy G. Exulceratio simplex. Bull Acad Med 1898; , Bolton C. Ulcer of the stomach. London: Edward Arnold, Meulengracht E. Fifteen years experience with free feeding of

14 608 SWAI ET AL. GASTROETEROLOG Vol. 90, o.3 patients with bleeding peptic ulcer: fatal cases. Arch Intern Med 1947;80: Osborn GR. The pathology of gastric arteries with special reference to fatal haemorrhage from peptic ulcer. Br J Surg 1954;41 : Chalmers TC, Zamcheck, Curtens GW, White FW. Fatal gastrointestinal hemorrhage. Clinicopathological correlations in 101 patients. Am J Clin Pathol 1953;22: Cornil AV, Ranvier LA. Manuel d'histologie pathologique. Vol. 2. Paris: Germer-Balliere & Cie, 1882: Hurst AF, Stewart MJ. Gastric and duodenal ulcer. Humphrey Milord, Oxford: Oxford Medical Publications, University Press, Anderson WAD. Pathology. London: Henry Kempton, Burger M. Altern und Krankheit. Leipzig: Georg Thieme, 1947: Jones FA. Modern trends in gastroenterology. London: Butterworth, orthfield TC. Factors predisposing to recurrent haemorrhage after acute gastrointestinal bleeding. Br Med J 1971;1: Rokitansky CFv. Handbuch der allgemeinen pathologischen Anatomie. Vol. 3. Vienna: Braunmuller and Seidel, 1846: For English translation see Rokitansky CFv. A manual of pathological anatomy. Vol. 2. London: Sydenham Society, 1854:33-4 (translated from German by Edward Sieveking).