It is ten at night in the medical emergency room. A 58-year-old man has passed black stools since the morning and fainted once in the bathroom. He had a coronary stent placed eight months ago and takes aspirin and clopidogrel. His pulse is 104, his systolic pressure 112 mmHg. The laboratory report says blood urea 62 mg/dL, creatinine 1.0 mg/dL, haemoglobin 10.8 g/dL. This is an illustrative case, not a real patient.
The Upper GI Bleed Pathway tells you what to do for him tonight: his Glasgow-Blatchford score, where he should be, when he should be scoped, what to give and what to hold. It no longer tells you why. The trials, the study figures and the history used to sit in its Protocols and Evidence tabs. They have moved here, so that the tool stays quick at the bedside. By the end of this article you will know where each of those instructions came from, and how far each one can be trusted.
Use this at the bedside
Open the tool: Upper GI Bleed Pathway
Get the CDSS app: install the MEDiscuss CDSS on your phone
1. What the Glasgow-Blatchford score was built to answer
Start with the question the score asks. It is not “how sick is this bleed”. It is “does this patient need anything a hospital provides”. The 2000 Lancet score was fitted to a composite of transfusion, endoscopic or surgical intervention, rebleeding and death, from data available before anyone had picked up an endoscope (Blatchford 2000). That is why it holds no age, no endoscopic finding and no comorbidity beyond the two that change haemodynamic reserve.
How well does it do that job? Area under the curve 0.86 to 0.92 for that composite in every large validation, and 0.64 for mortality alone in the 2017 international cohort (Stanley 2017). An area under the curve of 0.5 is a coin toss and 1.0 is perfect, so the score is good at its own question and poor at the other one.
“Then why is urea in it at all? Is it a kidney test?” Here it is not. A litre of blood in the small bowel is a large protein meal: the nitrogen is absorbed and converted to urea in the liver within hours while creatinine stays where it was. So a raised urea with a normal creatinine is a measure of how much blood has passed through the gut. Our patient’s urea of 62 mg/dL is about 10.3 mmol/L once divided by 6.006, which puts him in the 10.0 to 24.9 band, worth 4 points. His creatinine is normal. The urea is telling you about the bleed, not about his kidneys.
This is where the common mistake waits, and it is an easy one to make. If his report had printed BUN instead, and you had read the BUN as a urea, you would have placed him one or two bands too low. The tool asks which one you have and converts.
2. How the discharge threshold moved from 0 to 1
The original paper offered the score; it did not tell anyone to send a patient home. That step came later, and the sources have still not converged. This is the history, as the tool used to carry it.
| Source | Threshold, and the data behind it |
|---|---|
| Stanley 2009, four UK hospitals | Score 0. Sixteen per cent of 676 presentations scored 0, and a non-admission protocol at two of the hospitals cut the admission rate from 96 to 71 per cent with no adverse events in the patients managed at home |
| NICE CG141, 2012 | Score 0. Not revised since |
| International cohort 2017, 3,012 consecutive patients, six countries | Score 1 or less keeps a sensitivity of 98.6 per cent for survival without intervention and classifies 19 per cent of patients as low risk rather than 9 per cent. Among the 564 patients at 0 or 1, two died within 30 days and eight had endoscopic therapy |
| ESGE, ACG and the BSG bundle, 2019 to 2021 | Score 1 or less |
So the 2017 study is the reason ESGE, ACG and the BSG bundle accept a score of 1, while NICE still says 0. In the 2017 international validation a score of 1 or less had a sensitivity of 98.6 per cent for survival without intervention, and in the 564 patients so classified 1.8 per cent were transfused, 1.4 per cent had endoscopic therapy and 0.4 per cent died (Stanley 2017). Small numbers, but not zero. That is why the tool prints both thresholds at a score of exactly 1, and why a low score is only half of a discharge decision: the other half is whether the patient can come back.
Our patient does not get near this question. His urea gives 4 points, his haemoglobin of 10.8 g/dL another 3, his pulse 1, the melaena 1 and the faint 2: 11 in all. He is staying.
3. Three scores, three questions
Residents often ask which of the three scores is the best. The honest answer is that they were built for different jobs. The table below shows where each came from and what it was fitted to. The figures in brackets are the areas under the curve from the 2017 international cohort.
| Score | Origin, and what it was fitted to | What it answers here |
|---|---|---|
| Glasgow-Blatchford | 2000 Lancet paper; a composite of intervention, rebleeding and death, from pre-endoscopy data | Who needs the hospital, and how soon (0.75 for endoscopic therapy, 0.64 for mortality) |
| AIMS65 | A 187-hospital administrative database in the United States, fitted to in-hospital mortality by recursive partitioning. Two of its five items, albumin and INR, describe hepatic synthetic function and chronic illness rather than the bleed | Who has a high chance of dying and needs a higher level of care (0.77 for mortality, 0.62 for endoscopic therapy) |
| Rockall | The 1993 UK national audit, also fitted to mortality. The pre-endoscopy half is age, shock and comorbidity; the full form adds what the endoscopist saw | Admission form flags the patient at risk of dying. The full form, once available, describes the rebleeding risk that decides how long to keep them |
The Rockall score is the oldest of the three. It was derived from 4,185 cases in the 1993 UK national audit (Rockall 1996). AIMS65 came from a 29,222-patient derivation set in the United States (Saltzman 2011). Neither was built to tell you who can go home, and the Glasgow-Blatchford score was not built to tell you who will die. Use each for its own question.
4. Less blood is better: the Barcelona trial
“He is bleeding. Surely more blood is safer?” It feels that way. Restrictive transfusion is a treatment, not a saving. The 2013 Barcelona trial randomised 921 patients with severe upper GI bleeding to a transfusion trigger of 7 g/dL or 9 g/dL (Villanueva 2013).
- The restrictive group received fewer transfusions, as expected, but they also rebled less, 10 against 16 per cent, and more of them were alive at six weeks, 95 against 91 per cent.
- The mechanism is visible in the same paper: the hepatic venous pressure gradient rose in the liberal group within five days and did not in the restrictive group. Each unit of blood restores volume into a portal system whose pressure is what drove the bleed.
- The survival gain was concentrated in cirrhosis of Child-Pugh class A and B; there was none in class C.
Our patient has a coronary stent, so his trigger is not 7. The trials mostly excluded ischaemic heart disease, and that is why the tool raises the trigger to 8 g/dL for him. Here is the transfusion table whole, with the sources the tool now keeps in its reference list rather than in a fourth column.
| Patient | Transfuse red cells when | Target after transfusion | Source |
|---|---|---|---|
| Stable, no cardiovascular disease | Haemoglobin below 7 g/dL | 7 to 9 g/dL (ESGE); 7 to 10 g/dL (BSG) | Villanueva 2013; ESGE 2021, strong; moderate. ACG 2021. BSG 2020, strong; high |
| Stable, with cardiovascular disease | Haemoglobin 8 g/dL or below | 10 g/dL or above | ESGE 2021, strong; low. BSG 2020: consider a higher trigger with ischaemic heart disease |
| Cirrhosis with suspected variceal bleeding | Haemoglobin below 7 g/dL | 7 to 8 g/dL. Over-transfusion raises portal pressure and rebleeding; the survival gain in the Barcelona trial was in Child-Pugh A and B, not C | Baveno VII 2022, statement 6.2; Villanueva 2013 |
| Exsanguinating, unstable despite fluids | On the haemodynamics, not the haemoglobin, which lags the loss | Major haemorrhage protocol; platelets if actively bleeding with a count of 50 or below | BSG 2020, weak; very low for the platelet threshold |
5. Tranexamic acid: a null result that changed practice
Tranexamic acid saves lives in trauma and in postpartum haemorrhage, so it is natural to reach for it here. It does not work here. HALT-IT enrolled 12,009 patients in 15 countries, India among them, and gave a 1 g load followed by 3 g over 24 hours or placebo (HALT-IT 2020). Death from bleeding within five days was 4 per cent in both arms, and venous thromboembolism 0.8 per cent against 0.4 per cent, a relative risk of 1.85. ESGE and Baveno VII adopted that within a year (ESGE 2021; Baveno VII 2022).
“But surely it does no harm to try?” It does: venous thromboembolism nearly doubled. A resident who gives it to a bleeding ulcer today is treating on the strength of small trials that the large one overturned.
6. Sooner is not better: the timing of endoscopy
The standard is endoscopy within 24 hours, after resuscitation. Why not at once, if he is sick?
- The Hong Kong trial randomised 516 high-risk patients, Glasgow-Blatchford 12 or more, to endoscopy within 6 hours or between 6 and 24 hours of the gastroenterology consultation (Lau 2020). Thirty-day mortality was 8.9 per cent with urgent endoscopy and 6.6 per cent with early endoscopy, and further bleeding 10.9 against 7.8 per cent: neither difference significant, both in the wrong direction.
- The urgent group had more ulcers with active bleeding or visible vessels at the index endoscopy, and more endoscopic therapy. The extra hours let the acid-suppressed ulcer settle and let the resuscitation finish.
- Danish registry data, quoted in the BSG bundle, found mortality highest at both extremes of timing in the sickest patients (BSG 2020).
The tool keeps the two exceptions: the patient who stays unstable despite resuscitation, and suspected variceal bleeding, where Baveno VII asks for 12 hours.
7. Aspirin stays, and the evidence for it
Our patient is on aspirin for a stent. Should it stop? In the trial the BSG bundle cites, patients with ulcer bleeding on low-dose aspirin for secondary prevention were randomised after endoscopic haemostasis to continue the aspirin or to stop it for eight weeks. Continuation doubled recurrent bleeding, 10.3 against 5 per cent, and cut all-cause mortality from 12.9 to 1.3 per cent, because what killed the patients who stopped was not the ulcer. So his aspirin continues. His clopidogrel is held, with a restart date and the cardiologist in the conversation, because a stent placed eight months ago can thrombose. The tool prints exactly that.
8. The drugs, whole
The tool now shows its drug table in two halves on a phone: the drug, when and the dose in one; the source and the caution in the other. Here it is whole, as the tool carried it on 4 November 2026. Read the last column: it is where the cautions are.
| Drug | When | Dose | Source and caution |
|---|---|---|---|
| Proton pump inhibitor, before endoscopy | May be considered in any acute upper GI bleed; must not delay endoscopy | High-dose intravenous, the same regimen as after haemostasis | ESGE 2026, conditional; low. ACG 2021 makes no recommendation for or against. NICE CG141 (2012) said not to offer it before endoscopy. Three positions; the tool prints ESGE’s and says so |
| Proton pump inhibitor, after haemostasis | Forrest Ia, Ib or IIa treated endoscopically, and IIb whether treated or not | 80 mg intravenous bolus, then 8 mg per hour for 72 hours. Alternative: high-dose intravenous bolus twice daily, or high-dose oral twice daily. Then oral twice daily for 2 weeks, then once daily for a duration set by the lesion | ESGE 2021, strong; high. ACG 2021: 3 days high dose, then twice daily oral for 2 weeks |
| Erythromycin | Clinically severe or ongoing active bleeding, to clear the stomach before endoscopy | 250 mg intravenously, 30 to 120 minutes before endoscopy, single dose | ESGE 2021, strong; high. Baveno VII 6.19. Check the QT interval first: corrected QT calculator |
| Metoclopramide | The same indication, where intravenous erythromycin is not available | Intravenous; the guideline names no dose, so follow the local protocol | ESGE 2026, conditional; low. New in the 2026 update |
| Terlipressin | From the moment cirrhosis or variceal bleeding is suspected, before endoscopy | 2 mg intravenously every 6 hours, continued for 2 to 5 days or until haemostasis | BSG 2020 bundle, strong; high. Baveno VII 6.5 for the duration. Caution in ischaemic heart disease and peripheral vascular disease; ECG before starting in patients over 65; monitor sodium (Baveno VII 6.6) |
| Somatostatin or octreotide | Alternatives to terlipressin with equivalent outcomes in a 780-patient trial | Not dosed on this page; follow the local protocol | Baveno VII 6.5 names all three; BSG 2020 cites the non-inferiority trial |
| Antibiotic prophylaxis in cirrhosis | From admission, every patient with cirrhosis and upper GI bleeding, whatever the source turns out to be | Ceftriaxone 1 g intravenously once every 24 hours, in accordance with local resistance patterns and antimicrobial policy | Baveno VII 6.7 and 6.9. BSG 2020, strong; high. Duration up to 7 days in the ceftriaxone trial the bundle cites |
| Prothrombin complex concentrate | Warfarin patient who is unstable or bleeding severely | Weight and INR based; dose with haematology and the product insert, not from this page | ESGE 2021: low-dose vitamin K plus intravenous PCC if unstable. ACG-CAG 2022: PCC rather than fresh frozen plasma; suggests against vitamin K and against FFP |
| Tranexamic acid | Not indicated | None | HALT-IT 2020, 12,009 patients: no reduction in death from bleeding (4 per cent in both arms), venous thromboembolism nearly doubled (RR 1.85). ESGE 2021, strong; high, against. Baveno VII 6.38, against in variceal bleeding |
One line in the variceal part of the tool used to carry a figure. Terlipressin lowers portal pressure and reduces death by about a third against placebo in the Cochrane pooling. The dose and every caution are still in the tool.
9. The Forrest classes, whole
Forrest, Finlayson and Shearman described the appearances in the Lancet in 1974, when fibreoptic endoscopy of the bleeding patient was itself new (Forrest 1974). Half a century later the classes still decide who gets endoscopic therapy and who gets high-dose acid suppression. The tool splits this table in two; here it is in one piece.
| Class | Appearance | Endoscopic therapy (ESGE 2021 and 2026) | PPI afterwards |
|---|---|---|---|
| Ia | Spurting arterial bleeding | Adrenaline injection plus a second modality (thermal or mechanical), or an over-the-scope clip as monotherapy | High dose, 72 hours |
| Ib | Oozing | As Ia | High dose, 72 hours |
| IIa | Non-bleeding visible vessel | Thermal, mechanical or sclerosant, alone or with adrenaline; over-the-scope clip as an alternative | High dose, 72 hours |
| IIb | Adherent clot | Clot removal and treatment of what lies beneath, where the endoscopist is competent to manage the conversion to a higher-risk lesion (ESGE 2026) | High dose, 72 hours, treated or not |
| IIc | Flat pigmented spot | None | Oral, standard dose |
| III | Clean base | None | Oral, standard dose |
10. Where the guidelines differ
When two guidelines disagree, the tool prints both and names the one it follows. This is the full comparison it works from.
| Question | NICE CG141 (2012, updated 2016) | ESGE (2021, updated 2026) | ACG 2021, and ACG-CAG 2022 for antithrombotics |
|---|---|---|---|
| Discharge threshold | Glasgow-Blatchford 0, or pre-endoscopy Rockall 0 | Glasgow-Blatchford 1 or less | Glasgow-Blatchford 0 or 1 |
| PPI before endoscopy | Do not offer | May be considered; must not delay endoscopy | No recommendation for or against |
| Erythromycin before endoscopy | Not addressed | Yes, in severe or active bleeding | Suggested |
| Endoscopy timing | Immediately after resuscitation if unstable; within 24 hours for everyone else | Within 24 hours; not within 6 or 12 hours unless unstable despite resuscitation | Within 24 hours |
| DOAC reversal in severe bleeding | Not addressed | Consider a reversal agent or PCC | Suggests against PCC, idarucizumab and andexanet |
| Warfarin reversal | PCC for the actively bleeding patient on warfarin | Low-dose vitamin K plus PCC if unstable | PCC rather than FFP if reversal is needed; suggests against vitamin K and FFP |
| Tranexamic acid | Not addressed in the 2012 text | Against | Not recommended, on HALT-IT |
11. The Indian series: varices first
Every score here was derived in cohorts that were mostly non-variceal, and Indian cohorts are mostly variceal. Look at the variceal share in each series below, and at what the Kozhikode group found when it checked calibration and not only discrimination. Discrimination is whether the score ranks patients in the right order. Calibration is whether its predicted risk matches what actually happens. A score can do the first and fail the second.
| Series | Variceal share, and what the series found |
|---|---|
| Kozhikode 2023 | 49 per cent of endoscopic diagnoses. This series plotted calibration as well as discrimination: Glasgow-Blatchford and its modified form better than AIMS65 for transfusion and rebleeding, AIMS65 better for in-hospital mortality (15.4 per cent in that series), all three poor at predicting the need for endoscopic therapy, and every score poorly calibrated. A score of 1 or less was still not associated with adverse events, so the outpatient rule survives |
| PGIMER Chandigarh 2021 | 63.6 per cent |
| AIIMS Rishikesh 2023 | 52.9 per cent |
| Delhi 2019 | Each score markedly worse in variceal than in non-variceal bleeding |
| Kolkata 2025 | Discriminant cut-off for a composite adverse outcome at a Glasgow-Blatchford Score of 7 or more, the same threshold the international study found for endoscopic therapy. Proposes a four-item score of its own, which awaits external validation |
That is why the tool uses the score to send the low-risk patient home, opens the variceal pathway early from the liver disease box, and never quotes a Western mortality figure to a family in an Indian unit.
Back to the emergency room
Our patient scores 11, so he is admitted. His urea is a sign of blood in the gut, not of failing kidneys. He is transfused only if his haemoglobin falls to 8 g/dL, because of his heart. He gets no tranexamic acid. He is scoped within 24 hours, once resuscitated, not in the middle of the night. His aspirin continues; his clopidogrel is held with a written restart plan. Each of those steps is a line in the tool. Now you know what stands behind each line.
Try this
A 34-year-old woman had one episode of coffee-ground vomiting after a night of retching. Pulse 84, systolic 128 mmHg, urea 30 mg/dL, haemoglobin 13.1 g/dL, no melaena, no faint, no liver or heart disease. What is her Glasgow-Blatchford score, and may she go home? Illustrative case.
Answer. Her urea is 30 divided by 6.006, about 5.0 mmol/L, below the first band of 6.5. Every other item is normal for a woman. Her score is 0. ESGE, ACG, the BSG bundle and NICE all accept a score of 0 for outpatient care. She may go home with early outpatient endoscopy, if she can come back, and if somebody has told her what to do at the first black stool.
Take-home points
- The Glasgow-Blatchford score finds the patient who can go home. It is poor at finding the patient who will die.
- Convert urea from mg/dL to mmol/L, and never read a BUN as a urea.
- Transfuse at 7 g/dL, or at 8 g/dL with heart disease. More blood is not safer.
- Do not give tranexamic acid for gastrointestinal bleeding.
- Scope within 24 hours, after resuscitation. Keep aspirin taken for secondary prevention.
Use this at the bedside
Open the tool: Upper GI Bleed Pathway
Get the CDSS app: install the MEDiscuss CDSS on your phone
Declaration: the author built and maintains the tool. It is free to use. The figures in this article are those the tool carried in its Protocols and Evidence tabs on 4 November 2026, each from the source named beside it. The Upper GI Bleed Pathway is still in administrator preview, pending clinical review.
References
- Blatchford O, Murray WR, Blatchford M. A risk score to predict need for treatment for upper-gastrointestinal haemorrhage. Lancet. 2000;356(9238):1318-1321.
- Rockall TA, Logan RF, Devlin HB, Northfield TC. Risk assessment after acute upper gastrointestinal haemorrhage. Gut. 1996;38(3):316-321.
- Forrest JA, Finlayson ND, Shearman DJ. Endoscopy in gastrointestinal bleeding. Lancet. 1974;2(7877):394-397.
- Saltzman JR, Tabak YP, Hyett BH, Sun X, Travis AC, Johannes RS. A simple risk score accurately predicts in-hospital mortality, length of stay, and cost in acute upper GI bleeding. Gastrointest Endosc. 2011;74(6):1215-1224.
- Stanley AJ, Ashley D, Dalton HR, et al. Outpatient management of patients with low-risk upper-gastrointestinal haemorrhage: multicentre validation and prospective evaluation. Lancet. 2009;373(9657):42-47.
- Stanley AJ, Laine L, Dalton HR, et al. Comparison of risk scoring systems for patients presenting with upper gastrointestinal bleeding: international multicentre prospective study. BMJ. 2017;356:i6432.
- Villanueva C, Colomo A, Bosch A, et al. Transfusion strategies for acute upper gastrointestinal bleeding. N Engl J Med. 2013;368(1):11-21.
- Lau JYW, Yu Y, Tang RSY, et al. Timing of endoscopy for acute upper gastrointestinal bleeding. N Engl J Med. 2020;382(14):1299-1308.
- HALT-IT Trial Collaborators. Effects of a high-dose 24-h infusion of tranexamic acid on death and thromboembolic events in patients with acute gastrointestinal bleeding (HALT-IT): an international randomised, double-blind, placebo-controlled trial. Lancet. 2020;395(10241):1927-1936.
- Gralnek IM, Morris J, Laursen SB, et al. Endoscopic diagnosis and management of peptic ulcer bleeding: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Update 2026. Endoscopy. 2026;58:899-924.
- Gralnek IM, Stanley AJ, Morris AJ, et al. Endoscopic diagnosis and management of nonvariceal upper gastrointestinal hemorrhage (NVUGIH): European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Update 2021. Endoscopy. 2021;53(3):300-332.
- Laine L, Barkun AN, Saltzman JR, Martel M, Leontiadis GI. ACG Clinical Guideline: Upper Gastrointestinal and Ulcer Bleeding. Am J Gastroenterol. 2021;116(5):899-917.
- Siau K, Hearnshaw S, Stanley AJ, et al. British Society of Gastroenterology (BSG)-led multisociety consensus care bundle for the early clinical management of acute upper gastrointestinal bleeding. Frontline Gastroenterol. 2020;11(4):311-323.
- Abraham NS, Barkun AN, Sauer BG, et al. American College of Gastroenterology-Canadian Association of Gastroenterology Clinical Practice Guideline: Management of Anticoagulants and Antiplatelets During Acute Gastrointestinal Bleeding and the Periendoscopic Period. Am J Gastroenterol. 2022;117(4):542-558.
- de Franchis R, Bosch J, Garcia-Tsao G, Reiberger T, Ripoll C; Baveno VII Faculty. Baveno VII: Renewing consensus in portal hypertension. J Hepatol. 2022;76(4):959-974.
- Barkun AN, Almadi M, Kuipers EJ, et al. Management of Nonvariceal Upper Gastrointestinal Bleeding: Guideline Recommendations From the International Consensus Group. Ann Intern Med. 2019;171(11):805-822.
- Barkun AN, Laine L, Leontiadis GI, et al. Canadian Association of Gastroenterology Clinical Practice Guideline for the Endoscopic Management of Nonvariceal Nonpeptic Ulcer Upper Gastrointestinal Bleeding. Gastroenterology. 2025;169(5):863-891.
- National Institute for Health and Care Excellence. Acute upper gastrointestinal bleeding in over 16s: management. NICE guideline CG141. 2012, updated 2016.
- Akhila Arya PV, Thulaseedharan NK, Raj R, Unnikrishnan DC, Jacob A. AIMS65, Glasgow-Blatchford bleeding score and modified Glasgow-Blatchford bleeding score in predicting outcomes of upper gastrointestinal bleeding: An accuracy and calibration study. Indian J Gastroenterol. 2023;42(4):496-504.
- Sachan A, Dhibar DP, Bhalla A, Prakash A, Taneja S, Sharma V. Comparison of non-endoscopic scores for the prediction of outcomes in patients of upper gastrointestinal bleed in an emergency of a tertiary care referral hospital: a prospective cohort study. Arq Gastroenterol. 2021;58(4):534-540.
- Raj A, Kaeley N, Prasad H, et al. Prospective observational study on clinical and epidemiological profile of adult patients presenting to the emergency department with suspected upper gastrointestinal bleed. BMC Emerg Med. 2023;23(1):107.
- Rout G, Sharma S, Gunjan D, Kedia S, Nayak B, Shalimar. Comparison of various prognostic scores in variceal and non-variceal upper gastrointestinal bleeding: a prospective cohort study. Indian J Gastroenterol. 2019;38(2):158-166.
- Ganguly A, Samanta T, Das P, et al. Comparison of risk assessment scores of upper gastrointestinal bleeding: Proposal for a simplified score. Indian J Gastroenterol. 2025; published online 29 December 2025.
Leave a Reply