Acute gastrointestinal bleeding: upper GI

"Acute gastrointestinal bleeding" is a Level 1 topic from Section 2.1.6 in the first edition of the CICM Syllabus for the Second Part Examination, where the college does not offer any further granularity with respect to upper or lower GI. This is not an oversight, but rather the recognition that all forms of massive bleeding are relevant, whether they are above or below the ligament of Treitz. For the purposes of revision the distinction is still probably beneficial as very different approaches are required for each. This chapter deals with things that go wrong in the lumen of the upper gastrointestinal tract.  Historically, it has appeared in several SAQs:

• Question 9 from the second paper of 2024 (except they only specified "upper GI" in the examiner comments, not the stem)
• Question 3 from the first paper of 2024 (haematemesis)
• Question 1 from the first paper of 2017, which was all about the causes and management of a massive variceal haemorrhage
•  Question 1a and Question 1b from the second paper of 2001, where the college offered us an exsanguinating alcoholic. The real fun began when in the last part of the SAQ the wife asked whether he was going to survive (see Staging and prognosis of chronic liver disease in ICU)

In summary:

The best references for this would probably have to be large society guidelines, rather than review articles. 

• 2020 British (BSG) guidelines for acute upper GI bleeding
• 2021 American (ACG) guidelines for acute upper GI bleeding

Causes of gastrointestinal bleeding

Question 1 from the first paper of 2017 asked for four causes of massive upper GI haemorrhage, of which there is a long possible list. Here is a generic list incorporating upper and lower (and either) causes.  The list offered here was generated mainly using Oh's Manual, but contains conditions which are not listed in the canonic Chapter 42  (pp. 487,  "Acute  gastrointestinal  bleeding"  by Joseph  JY  Sung).

The examiners may expect the candidates to produce something more specific to the context of the SAQ stem, in which case there will probably be a need to tailor the structure of the responses to the presentation rather than putting them into the anatomical framework listed above. For the haematemesis-related Question 3 from the first paper of 2024, this would include the following etiologies, with incidence values pulled from Gibson & Odze (2011)

This was assembled using the familiar "VINDICATE" mnemonic device, but other methods are equally effective, as long as one always applies the same one to get the maximum benefit.

Assessment of upper GI bleeding

History

• For aetiology: liver disease, alcoholism, IVDU, Hep C, NSAID use, anticoagulation, weight loss, 
• For risk stratification (i.e. who needs endoscopy)
  • Syncope (indicates haemodynamic instability)
  • Frequency of haematemesis (indicates how fast the stomach is filling with blood)
  • Hypotension and tachycardia
  • Transfusion requirements in excess of 4 units of PRBCs over 12 hrs
  • Age over 60
  • Comorbidities, especially liver and cardiac failure

Examination

• For resuscitation:
  • Clinical features of shock that can act as endpoints, i.e vital signs, urine output
• For risk stratification and anticipation:
  • Airway grade
  • Access difficulty
• For aetiology:
  • Stigmata of chronic liver disease or IV drug use
  • Weight and abdominal masses, organomegaly, ascites
  • Lymphadenopathy

Investigations

• ABG for lactate, Hb and ionised calcium
• EUC for urea to observe the digestion of blood
• LFTs for albumin and other biomarkers suggestive of liver disease
• FBC, coags and TEG to see whether factors and platelets are required
• Group and hold, crossmatch (to prepare for transfusions)
• CXR, for gas under the diaphragm (perforated DU)
• CT mesenteric angiogram to look for embolisable lesions

Some of these plug in to calculate the Glasgow-Blatchford Bleeding Score (Laursen et al, 2012) or the AIMS65 score (Hyett et al, 2013) to help stratify these patients into risk groups and decide which require endoscopy and which do not.

Value of endoscopy in upper GI bleeding

Endoscopy is best. Why?

• Endoscopically, it is easy to visualise the source of bleeding and treat it simultaneously
• If one can visualise the source of bleeding, one can predict the risk of rebleeding:
  • Obvious bleeder: 85-90% risk of rebleeding
  • Obvious vessel: 35-55% risk of rebleeding
  • Clot: 30-40% risk of rebleeding
  • Reddish spot: 5-10% risk of rebleeding
  • Nothing found: 5% risk of rebleeding

Generally, postero-inferior duodenal wall ulcers and high lesser curve of stomach ulcers tend to rebleed most vigorously, owing to the presence of large arteries nearby. 

If endoscopy fails, or there is evidence of perforation, the next move may be either surgery rangiography. Nobody seems to agree precisely when to make this call. In general, if you are taking a patient to theatre for an emergency salvage surgery after an upper GI bleed, their mortality will be around 15-20%. On the other hand, in order to catch an upper GI bleeder on an angiogram, the rate of bleeding should be greater than 0.5ml/min.

Definitive haemostasis in upper GI bleeding

For peptic ulcers:

• Adrenaline injection - cheap, easy to learn, and effective
• Heat coagulation - with the added risk of perforation
• Clipping - no risk of perforation, but technically difficult in some sites

All these methods seem roughly equivalent in their effectiveness.

It is generally thought that the use of all three methods on the same lesion gives the greatest reduction in rebleeding risk.

For varices:

• Haemostasis:
  • Sclerotherapy - easier in the oesophagus; may cause chest pain, fever and mediastinitis
  • Banding - no inflammatory reaction, and at least as effective as sclerotherapy
  • Balloon tamponade - for when sclerotherapy and banding has failed, as you wait for TIPS. It is generally said that optimal pharmacological therapy is better than inexperienced use of balloon tamponade.
• Reducing portal hypertension pharmacologically
  • Terlipressin is the drug of choice in decreasing portal venous congestion, and it has beenshown to decrease mortality in variceal bleeding
  • Octreotide is the next best choice
  • Non-selective beta blockade might be helpful to prevent further re-bleeding, but the jury is still out.
• Reducing portal hypertension invasively
  • TIPS procedure - the redistribution of blood from the portal circulation certainly reduces the presure in oesophageal varices, and prevents rebleeding. TIPS decreases the chances of treatment failure in refractory variceal bleeding (in one study, the probability of remaining bleed-free was 97% in the TIPS group and 50% in the pharmacotherapy group). However, it is a risky undertaking. It is typically reserved for patients who have failed all the other therapeutic approaches. By commiting to this course of action, you exhange sanity for hemostasis, and condemn the patient to a greatly increased risk of hepatic encephalopathy.

• Surgical control - this means either the resection of the bleeding varices, or the construction of some sort of surgical shunt, eg. the Warren distal splenorenal shunt. This has never been demonstrated to improve survival, and certainly exposes the patient to a host of complications, most notable of which is the greatly increased difficulty of any future liver transplant procedures.

Risk factors for re-bleeding from varices

Question 1 from the first paper of 2017 asked specifically for "clinical indicators for risk of re-bleeding", which is a fancy way of asking for a list of risk factors. A good paper by Augustine et al (2010) actually presents an entire table of studies which investigated the various prognostic indicators which predict rebleeding in acute variceal haemorrhage. Re-bleeding in the case of many of these was rolled together into "5-day failure", a sort of composite endpoint together with mortality, which it obviously has some sort of effect upon. This was remixed into the following list of predictive features:

• Uncontrolled bleeding
  • Ongoing acute bleeding, or failure to control bleeding at initial endoscopy ("unable to band all varices")
  • Delay in the procedure
  • Number of bands which were used - according to Xu et al (2011), more than 6 bands is a bad sign
• Severe liver disease
  • Severity of liver disease: Child-Pugh and MELD scores (even their individual components!)
  • A hepatic venous pressure gradient (HVPG) in excess of 20mmHg
  • Aetiology of cirrhosis (apparently some causes are associated with greater risk of rebleeding)
  • Portal vein thrombosis
• Severe initial haemorrhage
  • High transfusion needs
  • Shock state
• Endoscopic features
  • Red Colour Sign - "red wale markings, cherry red or hematocystic spots" (Kleber et al, 1991)
  • Gastric varices, as opposed to oesophageal ones (Park et al, 2015)
• Laboratory features
  • Haematocrit
  • Platelet count
  • Coagulopathy (prolonged PT)

Transfusion thresholds

Question 9 from the second paper of 2024 c onfused the trainees by asking them to critically evaluate the threshold for PRBC transfusion in upper GI bleeding, which may have seemed like there was some complex evidence framework specific to this condition. But there is none. "Good answers provided a broad overarching description", extrapolating data from other studies which may not have even had GI bleeders in them,  "to adapt these principles to the specified sub-population". Adapt what principles? This is far from easy, as both TRICC and TRISS had never enrolled any such patients, as an example (TRICC specifically excluded anyone with acute blood loss).

Some data does exist, but it would find its way into the usual circulation for CICM trainees, as the data mostly concerns patients who were not especially sick. As an example, Villanueva et al, 2013, who enrolled 921 patients with "severe acute upper gastrointestinal bleeding" into a trial of Hb 90 vs 70 g/L, found that fully half of the restrictive cohort never required any blood products at all. The mortality was better in the restrictive group, but the difference was by about 4% overall, and limited to the Child-Pugh A and B patients. The TRIGGER trial by Jairath et al (2015) went for 80 vs 100g Hb in 936 patients, but excluded those who were "exsanguinating", as defined by "features of shock" (SBP under 100 and HR over 100) and a blood transfusion within 2 hrs of presentation. Quite right, the battlescarred ICU trainee from the West of Sydney might sigh (for whom such patients would never make it into the ICU); it does not matter what you do with your blood products here, in either case they were always going to be fine. 

So, in summary, if the upper GI bleed patient stops bleeding and begins to resemble the rest of the ICU population, then logically it makes sense that they would also have similar transfusion requirements to the rest of the ICU population, and the findings of TRICC and TRISS will apply to them. But what of the actual GI bleeder, who occupies their proper place in your attention on the night shift, which is front and centre? Again, there is no literature specifically about this group, but that is the case with a lot of ICU pathology, and sometimes we have to improvise. To answer Question 9 with something nuanced, one may have to borrow from trauma literature about permissive hypotension and haemostatic resuscitation.  Singer et al (2024) incorporated this into their management recommendations without a second thought.

Strategy to transfuse a severe ongoing GI bleed

• Decision to transfuse PRBCs is made on the basis of clinical criteria.
• Haemodynamically unstable, prior to definitive management
  • Aim for SBP ~85-90 mmHg
  • Use blood products exclusively for volume resuscitation in a balanced proportion to prevent coagulopathy
  • Therapy should be guided by organ perfusion, response to intervention and risk of ongoing bleeding, rather than Hb thresholds
• Haemodynamically stable, pre or post definitive management
  • Aim for Hb > 70g/L, same as with other ICU patients
  • Hb 80-90 for patients with acute ischaemic heart disease

Prevention of rebleeding after endoscopy

For peptic ulcer disease:

• Proton pump inhibitors
  • and it seems that PPI infusion has no advantage over twice-daily dosing
• Histamine (H2) receptor blockers

Tranexamic acid plays no role - there is good evidence that it is ineffective.

In general, drugs can prevent rebleeding post endoscopy, but they usually play little role in massive haematemesis, and on their own are ineffective.

For bleeding varices:

• Early endoscopic control: delayed endoscopy increases re-bleeding risk (Chen et al, 2012)- ideally, the bleed should be controlled within the first 12 hours, as this is associated with the best mortality 
• Terlipressin is the drug of choice, and it has been shown to decrease mortality in variceal bleeding
• Octreotide is the next best choice
• Non-selective beta blockade might be helpful but the jury is still out.
• TIPS decreases the chances of treatment failure in refractory variceal bleeding (in one study, the probability of remaining bleed-free was 97% in the TIPS group and 50% in the pharmacotherapy group)
• Antibiotics: In the patients with chronic liver disease, antibiotic prophylaxis is indicated. Sepsis promotes the risk of variceal bleeding, and variceal bleeding promotes the risk of sepsis. Literature demonstrates a benefit from antibiotics in this setting (the usual course is 7 days).
• Tranexamic acid - mentioned by the college in their answer to Question 1 from the first paper of 2017.Tavakoli et al published on this in 2017- they did not find any difference in rebleeding rate, nor any other outcome variable for that matter. The whole thing is very 80s. However, as the college answers are definitive, the savvy candidate would need to include this potentially pointless therapy in their answer.
•  Proton pump inhibitors: again, PPI infusion probably has no advantage over twice-daily dosing
• Sucralfate is also mentioned by the college in their answer to Question 1 from the first paper of 2017. The "local anti-fibrinolytic effect" is seen more in patients who have had sclerotherapy and then go on to bleed from post-sclerotherapy ulcers (i.e. no longer varices, but still technically a rebleed). This was reported upon by Brooks (1995). The specific benefit seems to be the result of sucralfate counteracting the pro-fibrinolytic effect of ethanolamine oleate, the specific sclerosant agent widely used in the 1990s.