Relevance of surgery in patients with non-variceal upper gastrointestinal bleeding
Upper GI bleeding remains one of the most common emergencies with a substantial overall mortality rate of up to 30%. In severe ill patients, death does not occur due to failure of hemostasis, either medical or surgical, but mainly from comorbidities, treatment complications, and decreased tolerated blood loss. Management strategies have changed dramatically over the last two decades and include primarily endoscopic intervention in combination with acid-suppressive therapy and decrease in surgical intervention. Herein, we present one of the largest patient-based analysis assessing clinical parameters and outcome in patients undergoing endoscopy with an upper GI bleeding. Data were further analyzed to identify potential new risk factors and to investigate the role of surgery.
Patients and methods
In this retrospective study, we aimed to analyze outcome of patients with an UGIB and data were analyzed to identify potential new risk factors and the role of surgery. Data collection included demographic data, laboratory results, endoscopy reports, and details of management including blood administration, and surgery was carried out. Patient events were grouped and defined as “overall” events and “operated,” “non-operated,” and “operated and death” as well as “non-operated and death” where appropriate. Blatchford, clinical as well as complete Rockall-score analysis, risk stratification, and disease-related mortality rate were calculated for each group for comparison.
Overall, 253 patients were eligible for analysis: endoscopy was carried out in 96% of all patients, 17% needed surgical intervention after endoscopic failure of bleeding control due to persistent bleeding, and the remaining 4% of patients were subjected directly to surgery. The median length of stay to discharge was 26 days. Overall mortality was 22%; out of them, almost 5% were operated and died. Anticoagulation was associated with a high in-hospital mortality risk (23%) and was increased once patients were taken to surgery (43%). Patients taking steroids presented with a risk of death of 26%, once taken to surgery the risk increased to 80%. Patients with liver cirrhosis had a risk of death of 42%; we observed a better outcome for these patients once taken to theater. Clinically, once scored with Blatchford score, statistical correlation was found for initial need for blood transfusion and surgical intervention. Clinical as well as complete Rockall score revealed a correlation between need for blood transfusion as well as surgical intervention in addition with a decreased outcome with increasing Rockall scores. Risk factor analysis including comorbidity, drug administration, and anticoagulation therapy introduced the combination of tumor and non-steroidal antirheumatic medication as independent risk factors for increased disease-related mortality.
UGIB remains challenging and endoscopy is the first choice of intervention. Care must be taken once a patient is taking antirheumatic non-steroidal pain medication and suffers from cancer. In patients with presence of liver cirrhosis, an earlier surgical intervention may be considered, in particular for patients with recurrent bleeding. Embolization is not widely available and carries the risk of necrosis of the affected organ and should be restricted to a subgroup of patients not primarily eligible for surgery once endoscopy has failed. Taken together, an interdisciplinary approach including gastroenterologists as well as surgeons should be used once the patient is admitted to the hospital to define the best treatment option.
KeywordsUpper GI bleeding Outcome Surgery Scoring systems
Acute upper gastrointestinal bleeding (UGIB) is a frequent medical emergency with an incidence between 48 and 160/100000 per year [1, 2]. The first nationwide audit of patients with UGIB was carried out in the UK reporting an overall mortality rate of 14%, 11% for patients admitted with bleeding, and 33% for in-patients [3, 4, 5]. Generally, mortality rates are considered to be around 10–15%, which can be much higher once more critical ill patients with recurrent bleeding are taken into consideration [6, 7, 8, 9]. Other data indicate a recent mortality rate of under 15% during hospital readmission . More recent data of the following audit on UGIB in the UK over a decade later demarked a decreased overall mortality of 10% while inpatient mortality was still higher than mortality for patients admitted to the hospital for UGIB (7 versus 26%, respectively) . Consistently, mortality remains to be highest in patients with variceal bleeding and malignancies . However, the majority of deaths does not occur due to failure of hemostasis, either medical or surgical, but mainly from comorbidities, treatment complications as well as poorly tolerated blood loss in severe ill patients .
The management of upper GI bleeding is a rather complex one and requires a multidisciplinary approach involving surgeons, gastroenterologists, and interventional radiologists . Endoscopy remains the first treatment option until today [13, 14, 15]. Management strategies have changed over the last two decades and include primarily endoscopic evaluation within the first 24 h followed by hemostatic drug application or clipping resulting in combination with acid-suppressive therapy in a decrease in surgical intervention. Recently, at tertiary centers, radiological intervention may be offered for the management of upper GI bleeding. Various surgical procedures may be feasible for bleeding control once endoscopic therapy has failed including ulcera excision, partial or total gastric resection including Billroth I and II procedures as well as direct ligation of the arteria gastroduodenalis or even Whipple-procedure for duodenal ulcera. Hospitalization is mandatory at least for patients with cirrhosis and variceal bleeding but may also be favored for non-variceal bleeding due to a wide range of clinical severity and outcome in these patients [16, 17]. Several scoring systems have been designed and introduced in clinical practice to estimate the risk of recurrent bleeding, severe clinical complications, and death including the Blatchford as well as Rockall scoring system and more recently the Glasgow Blatchford bleeding score (GBS) [18, 19]. Blatchford and modified GBS is used to identify patients with UGIB based on clinical findings who may need clinical intervention before endoscopy once scored greater 0. Rockall score is divided into two different scores while patients with a clinical Rockall score greater than 0 are considered to be at high risk for severe outcome before endoscopy. Recently, a better outcome prediction for patients with suspected acute upper GI bleeding was seen in a prospective multicenter validation for the GBS compared with the Rockall scoring system . However, complete Rockall score includes endoscopic findings indicated once patients scored greater than 2 which meant a high risk for recurrent bleeding and death [4, 5, 16, 17, 20].
Our study focuses on patients admitted to a university hospital as well as in-patients with acute UGIB. Herein, we subgrouped patients according to the intervention carried out (endoscopy, surgery, mortality rate) as well as in-hospital mortality and patients were analyzed for clinical parameters, risk scoring systems as well as clinical outcome. As supported by our data, we emphasize the need for early surgical intervention in regard to the outcome and present a risk factor analysis including a binary decision tree for clinicians.
Methods and patients
In this retrospective study, data of patients with a UGIB were analyzed for outcome, to identify potential new risk factors and to investigate the role of surgery. An electronic search was made of all adult patients with an acute UGIB at the University Medical Center Goettingen (UMG, Goettingen, Germany)—a tertiary care, university hospital. The majority of patients were admitted as an emergency to our hospital. Retrospectively, data were collected by the comprehensive review of the original data records of all patients between January 2005 and December 2012. Data were collected according to the ClinData® clinical diagnosis system as used routinely in our hospital for patient administration management. Inclusion criteria were an upper GI bleeding followed by an endoscopic intervention and/or direct surgical intervention. Patients were included in the study regardless of any previous proton pump inhibitor (PPI) treatment. Patients were excluded, if they did not undergo endoscopy and/or operation, were 18 years old or younger, or bled from lower GI source. Patients with variceal bleeding were not further analyzed due to a significant higher known mortality rate and almost compulsory need for surgical intervention.
Patients were directly transferred to the operation room (OR) once endoscopic intervention was not possible and sufficient bleeding control has failed. In case of a re-bleeding event, patients were re-assessed and re-endoscopy was carried out. In case sufficient bleeding control was not achieved, patients were directly taken to the OR independently of the reason for bleeding including tumor bleeding. Insufficient bleeding control was defined as 6 or more units of red blood cell transfusions within 24 h or no endoscopic intervention (e.g., clipping) was possible as defined by our clinical standard.
Definitions and parameters
Patient events were grouped into five groups and were generally defined as “overall” events and “operated,” “non-operated,” and “operated and death” as well as “non-operated and death.” Demographic data, clinical details including the Forrest-classification for non-variceal bleeding, laboratory results, endoscopy reports, and details of management including blood administration, endoscopic therapy, and surgery details were collected and analyzed. Data analysis was used to calculate the potential risk for clinical intervention for each group according to the Blatchford score. Overall risk of death was calculated using the clinical Rockall score as well as the complete Rockall score according to the collected clinical data and endoscopic findings as described before. Outcome and disease-related mortality rate were calculated for each group for comparison. Postinterventional or postoperative outcome as well as risk factor analysis was carried out independent of any previous PPI treatment in patients with a UGIB undergoing endoscopy, surgery, or both.
Data are presented as percentages with numerator/denominator and as summary statistics of median ± ranges or mean ± SD as appropriate and were calculated using the R statistical computing environment (version 3.1, www.r-project.org). To test for clinical factors with risk scores, the R statistical software (version 3.1) was used as well. Associations to categorical variables and significance were computed using the Kruskall-Wallis test. Association to continuous variables is shown as scatter plots and a significance test for correlation via Kendall’s tau was carried out. A chi-square test was used for significance analysis once a correlation was found. A p value of less than 0.05 was considered statistically significant. A binary decision tree using variable selection to predict risk of death depending on relevant risk factors was computed using R library rpart.
Patients’ characteristics and clinical parameters
Operation and death
Hemoglobin concentration g/dl
NSAR + steroids
Reason for bleeding
Incidence of operation and overall as well as treatment-related mortality after endoscopic intervention in patients undergoing endoscopy for upper GI bleeding
Endoscopic intervention n (%)a
Operation and death
Suprarenin and clipping
Suprarenin and other
Correlation between anatomical side of bleeding, initial endoscopic treatment, need for operation, and mortality
Endoscopic treatment n (%)
Supra and clip
Supra and other
Bleeding of anastomosis
Stomach and duodenum
Esophagus and stomach
Overall, 47 (16.6%) patients were operated due to persistent bleeding (40 patients, 85.1%) or due to failure of endoscopic intervention (7 patients, 14.9%). In detail, ulcus excision was carried out in 11 cases, a Billroth I (BI) in 25, a Billroth II (BII) in 6, a total gastrectomy in 2 cases, and a primarily direct ligation of the arteria gastroduodenalis was performed in nine cases and a duodenotomy in two cases. Out of them, a reoperation due to missing bleeding control was carried out in six (2.1%) cases. In three patients, bleeding was ongoing after ulcus excision followed by BI resection and, in two cases, ligation of a. gastroduodenalis was insufficient for bleeding control followed by relaparotomy and partial gastrectomy; in one case, an anastomosal ulcus with bleeding was seen after gastroenterostomy after primary BI resection. In one patient, a radiological embolization was carried out followed by necrosis of the upper duodenal C and distal stomach and a Whipple procedure.
None of the patients presenting with UGIB was initially treated by angiography followed by an intervention. Patients were treated either via endoscopy, operation, or both as indicated above. In one case of recurrent bleeding, angiographic intervention was carried out, resulting in organ necrosis and surgical intervention as described above.
Duration of stay
Median length of stay was 26.3 days. Patients undergoing endoscopic treatment only spent 24 days in hospital and patients undergoing primary or secondary surgery were admitted for 37 days without any statistical difference.
Risk of death after surgery in different subgroups of patients
Risk of death
Risk of death
Risk of death
Out of all the patients undergoing surgical intervention, there was no statistical difference for adjusted mortality regarding type of procedure itself; neither did the type of endoscopic intervention. Once taken to the OR, we observed a trend toward better in-hospital disease-related survival compared to endoscopy alone, but findings did not reach significance. In-hospital mortality was not influenced by hemoglobin concentration at admission or substitution of erythrocyte suspension, neither before nor after intervention or surgery.
Overall, we noted 88 patients (34.8%) with recurrent bleeding and surgical intervention was necessary in 30.1% (27 patients) with an event of rebleeding and an overall mortality of 22.7% (20 patients) without any statistical difference between the groups (p = 0.76, 95%CI). Re-endoscopy was carried out in 87 cases (34.4%); out of them, 20 patients (22.9%) died resulting in a high in-hospital mortality rate of almost 40%. Only one attempt for re-endoscopy was carried out before further steps and consisted mainly in suprarenin application and clipping (n = 29), in clipping alone (n = 13), or in suprarenin application alone (n = 37). In one case, presence of a spontaneous perforation led to direct surgical intervention. Mortality rate of conservative treated patients (n = 14) without surgical intervention was 15.9%. Most patients with need for surgical intervention presented with either gastric or duodenal ulcera (89.3%). No correlation was seen between hemoglobin concentration at admission or substitution of erythrocyte suspension and outcome (data not shown), neither before nor after intervention or surgery for recurrent bleeding.
Clinical outcome and clinical scoring systems
Upper GI bleeding is a frequent emergency and remains with a substantial overall mortality rate of up to 30% a life-threatening situation. Endoscopy remains the first treatment of choice [13, 14, 15]. Recent advanced technical progress in endoscopic techniques such as clipping (e.g., OTSC clips) and 24-h ready-to-treat endoscopic facilities widely available also in non-urban centers and district hospitals have reduced mortality and the need for surgical intervention [21, 22]. In most cases, endoscopic intervention is needed and feasible. Therefore, surgical intervention has dramatically changed over the last 10 years, and indication of surgery is limited only for patients where endoscopy therapy has failed . Moreover, the role of surgery has recently been challenged for patients with acute peptic ulcer bleeding, stressing the role of conservative treatment or radiological intervention . Also, the role of radiological intervention will possibly increase once more centers with interventional radiology are available. General application of proton pump inhibitors before intervention showed no significant reduction of mortality rate, event of rebleeding, or need for surgery in a recent meta-analysis . But, a recent large British study named wide pharmacological therapy and endoscopic intervention as the main determining factors for reduced mortality and improved clinical outcome . However, at least for recurrent bleeding, once blood thinners are used like clopidogrel or ASS, pharmacological proton pump inhibitors should always be applied [25, 26].
As shown here, patients taking NSAR and presenting with a tumor bleeding were at a high risk for increased mortality. However, patients taking blood-thinners were not at a higher risk of recurrence with poor outcome, which is in line with recent findings not showing a decreased outcome following non-variceal upper gastrointestinal bleeding . Herein, we observed a probably higher overall mortality rate which is due to presentation of several severely ill patients which is typical for a tertiary center. Therefore, we also observed a higher rate for surgical intervention. Our data here indicated the need for surgical intervention in 30% for patients with an event of rebleeding with an unadjusted overall mortality of 23% for patients suffering under a non-variceal bleeding. Patients presenting with liver cirrhosis or renal insufficiency showed with a significant low in-hospital mortality once taken to surgery. At least in patients with cirrhosis, surgical intervention may also be considered earlier since cirrhosis itself is a risk factor for recurrent bleeding . In addition, prophylactic application of antibiotics in patients with liver cirrhosis improved outcome and reduced rebleeding events after primary endoscopic intervention and hemostasis . Bacterial infection in patients with liver cirrhosis was significantly higher in patients with higher Child-Pugh score, active alcoholism, and no antibiotic prophylaxis, and therefore, a careful evaluation of the prophylactic antibiotic use is necessary in patients with at least peptic ulcer bleeding . In-hospital mortality was at least in our study lower after surgery and in the majority of cases, a limited resection was sufficient and fast bleeding control was achieved indicating the importance of surgery for patients with upper GI bleeding.
Patients taking steroids had also in our study a significantly higher risk for surgical complications and rebleeding indicating a possible higher benefit for these patients for embolization techniques. First coil embolization techniques were performed in 1975 . In the following years, arterial interventions have improved significantly due to advances in catheter technology and improved embolization materials, which both resulted in fewer complications . There are two commonly known indications for angiography. The first one is that if the bleeding site cannot be identified by endoscopy or computed-tomoraphic angiography. The second indication is if the source of the bleeding cannot be controlled endoscopically . Angiography, however, is particularly indicated when it comes to high-risk surgical patients. In these cases, radiological intervention is favored over surgery as treatment of choice after failed endoscopic therapy, because it is minimally invasive and associated with lower mortality . Further, Ripoll et al. and Defreyne et al. were able to show that there is no significant difference in outcome between patients managed with surgery versus arterial embolization [35, 36]. For a proper embolization to control GI, hemorrhage interventional radiologists have three options: the pharmacologic control with use of vasopressin; embolization using temporary or permanent embolics, which may be either solid or liquid; and catheter-induced vasospasm . However, the abovementioned techniques require highly trained specialists with many years of experience, which usually can only be found in hospitals with maximum care facilities. Smaller hospitals in rural areas are likely to lack a radiology department with a dedicated interventional unit. But, the majority of hospitals up to date do not have radiological intervention possibilities and endoscopy and/or surgery remains the only option [21, 37, 38]. Furthermore, as also seen in one of our cases, embolization maintains the risk of necrosis of affected organs leading to emergency surgical procedures up to a pancreaticoduodenectomy and partial gastrectomy with highly increased postoperative morbidity and mortality.
Taken together, we believe management of UGIB remains challenging and endoscopy is the first choice of intervention. Particular care must be taken once a patient suffers from a cancer and/or is taking antirheumatic non-steroidal pain medication. In patients with presence of liver cirrhosis, an earlier surgical intervention may be considered once quick and efficient bleeding control is not achieved via endoscopy. In particular, once endoscopy has failed and interventional radiology is not available, surgery remains not only the last option but should be taken into consideration from the beginning. In contrary, patients taking steroids have a significantly higher risk of surgical complications and should possibly subjected to embolization, if available. In summary, our data stress the importance of a multidisciplinary approach once a patient is admitted to the hospital for non-variceal upper gastrointestinal bleeding event.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
This article does not contain any studies with human participants or animals performed by any of the authors.
- 2.Lewis JD, Bilker WB, Brensinger C, Farrar JT, Strom BL (2002) Hospitalization and mortality rates from peptic ulcer disease and GI bleeding in the 1990s: relationship to sales of nonsteroidal anti-inflammatory drugs and acid suppression medications. Am J Gastroenterol 97(10):2540–2549CrossRefPubMedGoogle Scholar
- 4.Rockall TA, Logan RF, Devlin HB, Northfield TC (1995a) Incidence of and mortality from acute upper gastrointestinal haemorrhage in the United Kingdom. Steering committee and members of the National Audit of acute upper gastrointestinal haemorrhage. BMJ 311(6999):222–226CrossRefPubMedPubMedCentralGoogle Scholar
- 10.Stromdahl M, Helgeson J and Kalaitzakis E (2016) "Emergency readmission following acute upper gastrointestinal bleeding." Eur J Gastroenterol HepatolGoogle Scholar
- 15.Sung JJ, Chan FK, Chen M, Ching JY, Ho KY, Kachintorn U, Kim N, Lau JY, Menon J, Rani AA, Reddy N, Sollano J, Sugano K, Tsoi KK, Wu CY, Yeomans N, Vakil N, Goh KL (2011) Asia-Pacific working group consensus on non-variceal upper gastrointestinal bleeding. Gut 60(9):1170–1177CrossRefPubMedGoogle Scholar
- 18.Aquarius M, Smeets FG, Konijn HW, Stassen PM, Keulen ET, Van Deursen CT, Masclee AA, Keulemans YC (2015) Prospective multicenter validation of the Glasgow Blatchford bleeding score in the management of patients with upper gastrointestinal hemorrhage presenting at an emergency department. Eur J Gastroenterol Hepatol 27(9):1011–1016CrossRefPubMedGoogle Scholar
- 19.Yang, H. M., S. W. Jeon, J. T. Jung, D. W. Lee, C. Y. Ha, K. S. Park, S. H. Lee, C. H. Yang, J. H. Park and Y. S. Park (2015). "Comparison of scoring systems for nonvariceal upper gastrointestinal bleeding: a multicenter prospective cohort study." J Gastroenterol Hepatol.Google Scholar
- 24.Sreedharan A, Martin J, Leontiadis GI, Dorward S, Howden CW, Forman D, Moayyedi P (2010) Proton pump inhibitor treatment initiated prior to endoscopic diagnosis in upper gastrointestinal bleeding. Cochrane Database Syst Rev 7:CD005415Google Scholar
- 27.Teles Sampaio E, Maia L, Salgueiro P, Marcos-Pinto R, Dinis-Ribeiro M and Pedroto I (2016) "Antiplatelet agents and/or anticoagulants are not associated with worse outcome following nonvariceal upper gastrointestinal bleeding." Rev Esp Enferm Dig 108Google Scholar
- 29.Kuo MT, Yang SC, Lu LS, Hsu CN, Kuo YH, Kuo CH, Liang CM, Kuo CM, Wu CK, Tai WC, Chuah SK (2015) Predicting risk factors for rebleeding, infections, mortality following peptic ulcer bleeding in patients with cirrhosis and the impact of antibiotics prophylaxis at different clinical stages of the disease. BMC Gastroenterol 15:61CrossRefPubMedPubMedCentralGoogle Scholar