Acute gastrointestinal bleeding

ACR AC Lower GI Tract Bleeding (2021 update); ACR AC Nonvariceal UGIB; ESGE 2021/2015

The whole study rests on catching contrast leaving the vessel while it is still leaving. That needs an arterial acquisition, an unenhanced series to prove the density is new, and a later venous acquisition to show it accumulating and moving — which is why a routine single portal-venous abdomen is the wrong study, not merely a suboptimal one.

Overt gastrointestinal bleeding with haemodynamic significance, typically after endoscopy has failed to identify or control the source, or where endoscopy is not feasible.

Referenced decision support — confirm against your local protocol.

Decision support, not a directive. Protocols and timings shown are typical published ones — your local protocol takes precedence, and the vetting radiologist decides.

The request in front of you

Everything is optional. Leave a field alone and the answer assumes nothing — the verdict updates as you go.

Study requested

What the referrer actually asked for. It is evidence of intent, not a constraint on the right answer.

Contrast as written

What the request form says, not what it should say.

Age

Pick a band, or type an exact age if it matters.

years

Pregnancy status

Haemodynamic state

The fact that lets a pathway waive its own requirements. A crashing patient does not wait for a score.

Where is the patient?

Some pathways change in the emergency department — an acutely threatened limb is not an elective work-up.

Renal risk factors

The question a vetter can actually answer from the request. An explicit “none known” is a real answer, and it removes checks rather than deferring them.

Previous contrast reaction

The pathway — tap anything already done

Marking a study complete moves the answer on. A patient arrives partway through a pathway far more often than at the start of one.

Accept as requested

CT Mesenteric Angiogram — Biphasic

CT Angiogram — Mesenteric

What we'd amend, and why

Active haemorrhage is diagnosed as contrast that is inside the bowel lumen and changes between acquisitions. The unenhanced series is what allows an intrinsically dense clot, a retained tablet or residual contrast from an earlier study to be excluded as the cause. The arterial acquisition catches the jet at the moment of extravasation and shows the feeding vessel for embolisation, and the later acquisition shows the collection enlarging or moving with peristalsis, which is what separates true extravasation from a hyperdense mucosal lesion. Positive oral contrast is prohibited because it is indistinguishable from the finding being sought.

Nothing needs resolving before this goes ahead. Routine checks below are owned downstream.

How was this decided?

pathwayadult — Adults
rulerule-contrast-reaction-premed — Prior contrast reaction and elective premedication; checked by Nurse before the scan
rulerule-metformin — Metformin and iodinated contrast; checked by Radiographer at the scanner
rulerule-paeds-dose — Child-sized technique and contrast dose; checked by Radiographer at the scanner
rulerule-pregnancy-ionising — Pregnancy status before an ionising exposure; checked by Radiographer at the scanner
rulerule-renal-iodinated — Kidney function and intravenous iodinated contrast; checked by Radiographer at the scanner
rulerule-iv-access — Intravenous access adequate for the planned injection; checked by Radiographer at the scanner

Decision support only. Local protocol takes precedence.

Handled at the scanner(4)nothing for you to do

Settled and owned downstream. Each returns to a radiologist only on the stated trigger.

Metformin and iodinated contrast
Confirm whether the patient takes metformin or a metformin-containing combination, and if so whether ACR Category II applies (eGFR below 30, known or suspected AKI, or an arterial catheter study likely to cause renal embolisation). If Category I — that is, no AKI and eGFR at or above 30 — no action of any kind is needed.
Radiographer at the scannerAt the scanner
Flags back if: The patient takes metformin AND meets ACR Category II — eGFR below 30 mL/min/1.73 m2, known or suspected acute kidney injury, or an arterial catheter procedure with likely renal arterial embolisation. Metformin plus a normal or mildly reduced eGFR is explicitly NOT a flag-back: there is no need to stop metformin before or after intravenous iodinated contrast in Category I patients, and no need to re-check creatinine afterwards.
Child-sized technique and contrast dose
Confirm that a size- or weight-based protocol is selected — child-sized kV and mAs against size-based diagnostic reference ranges — and that contrast volume is calculated by weight rather than taken from an adult default. Weight-based iodinated contrast volumes of roughly 1.5–2.0 mL/kg are widely used in paediatric CT.
Radiographer at the scannerAt the scanner
Flags back if: No paediatric or size-based protocol exists on the scanner for the requested examination, or the requested coverage or number of phases exceeds what the clinical question needs — for example a multiphase study where a single phase answers it, or whole-body coverage for a focal question.
Pregnancy status before an ionising exposure
Make the pregnancy enquiry immediately before the exposure and record the answer. In the UK this is a statutory operator duty discharged at the time of exposure under the employer’s written procedures required by IR(ME)R 2017 — it is not something the vetting radiologist can perform or pre-empt, and a request is complete without it.
Radiographer at the scannerAt the scanner
Flags back if: The patient states that she is, or may be, pregnant AND the uterus is in or near the primary beam. The exposure is then paused for re-justification by the IR(ME)R practitioner before it proceeds.
Intravenous access adequate for the planned injection
Site and test a cannula that supports the protocol flow rate, preferring an antecubital or large forearm vein, and observe the injection for extravasation. A 20-gauge or larger cannula is preferred for flow rates of 3 mL/s or more.
Radiographer at the scannerAt the scanner
Flags back if: No cannula can be sited that supports the protocol flow rate — for example only a 22-gauge hand or foot cannula for a CT angiogram needing 4–5 mL/s; or the only available access is a central line or port that is not labelled power-injectable; or an extravasation occurs.

Pathways

Big forks are separate pathways; the first whose conditions match is the one used.

Haemodynamically unstable — massive haemorrhage

Role	Study & protocol	Why this answers the question
First line	CT Angiogram — Mesenteric CT Mesenteric Angiogram — Biphasic usually appropriate	In the unstable patient the CT angiogram stops being a diagnostic test and becomes the map for the treatment that is about to follow, so it is arranged with the interventional team already alerted and the report is expected to name the feeding vessel rather than describe a territory. Everything about the protocol earns its place under time pressure: the unenhanced series is seconds and is what stops a dense clot or a retained tablet being called extravasation, the arterial acquisition is where the jet is seen, and the later venous acquisition is what proves the collection is growing. Two things are deliberately absent from this arm. Labelled red-cell scintigraphy is one — an acquisition of an hour and a half that returns a bleeding territory and no vessel is time this patient does not have, and requesting it here is the single commonest sequencing error on this card. Waiting for endoscopy is the other, where endoscopy has already failed or the patient is too unstable to tolerate it. Renal risk does not defer this scan either: the hypovolaemia is the larger threat to the kidney and the alternative to the study is not a safer study, it is exsanguination.

What follows a positive study here is embolisation, and what follows a negative one in a patient who is still bleeding is repeat endoscopy or surgery — not another acquisition. That is why this arm is one rung long.
A variceal bleed is a different pathway: the immediate treatment is endoscopic and pharmacological, and the imaging question is portal venous anatomy for a shunt rather than arterial extravasation.

Occult or obscure bleeding with no active haemorrhage

Role	Study & protocol	Why this answers the question
First line	CT Enterography CT Enterography — Neutral Oral with Enteric-Phase Acquisition usually appropriate	The target here is a lesion, not a leak. Iron-deficiency anaemia with a normal gastroscopy and colonoscopy means the source is most often small bowel — an angioectasia, a small bowel tumour, a Meckel diverticulum, a Dieulafoy lesion — and finding it needs the lumen distended with a neutral agent and an enteric-phase acquisition timed to peak mural enhancement, so that an enhancing mucosal lesion stands out against a low-attenuation lumen. Two honest caveats belong on the report. Endoscopy still leads the pathway: after a negative bidirectional endoscopy the usual next investigation for suspected small-bowel bleeding is capsule endoscopy, and CT enterography is complementary to it rather than a replacement — it is specifically the study to do FIRST where a stricture, previous surgery or Crohn disease makes capsule retention a real risk, and the study to do when capsule endoscopy is negative, incomplete or unavailable. And a CT angiogram is the wrong test for this patient: without active bleeding at the moment of the acquisition there is nothing to extravasate, so a negative result carries no information at all.
Second line	MRI Small Bowel (MR Enterography) MR Enterography — small bowel	The same anatomical question without ionising radiation, which matters when the patient is young, when the search is likely to be repeated over years, or when Crohn disease is part of the differential and mural activity has to be graded as well as located. It is second only because availability and the hour of staged drinking make it slower to arrange, not because it answers less.
Problem solving	CT Angiogram — Mesenteric CT Mesenteric Angiogram — Biphasic	The study returns the moment the bleeding does. If this patient presents again with overt haemorrhage, the question reverts to where the blood is leaving the vessel and the multiphase angiogram is the right acquisition — performed while the bleeding is happening, not booked for the next list.

This arm is about a patient who is not bleeding today. Overt bleeding in the same patient tomorrow is the default ladder, and the two should not be confused because the words on the request card are similar.

Adults

Matches your inputsDefault

Role	Study & protocol	Why this answers the question
First line	CT Angiogram — Mesenteric CT Mesenteric Angiogram — Biphasic usually appropriate	Active haemorrhage is diagnosed as contrast that is inside the bowel lumen and changes between acquisitions. The unenhanced series is what allows an intrinsically dense clot, a retained tablet or residual contrast from an earlier study to be excluded as the cause. The arterial acquisition catches the jet at the moment of extravasation and shows the feeding vessel for embolisation, and the later acquisition shows the collection enlarging or moving with peristalsis, which is what separates true extravasation from a hyperdense mucosal lesion. Positive oral contrast is prohibited because it is indistinguishable from the finding being sought.
Second line	Red-cell scintigraphy (Tc-99m labelled RBC scan) Labelled red-cell scintigraphy — dynamic abdominal acquisition usually appropriate	The sequencing point is sensitivity to rate. CT angiography needs enough contrast to leave the vessel within a single arterial acquisition, which puts its practical detection threshold around 0.3-0.5 mL/min; labelled red cells accumulate over a dynamic acquisition of an hour and a half and are described as detecting bleeding an order of magnitude slower, around 0.1 mL/min. So this is the study for the patient who is still bleeding after a negative CT angiogram, and for bleeding that is slow or intermittent rather than brisk — the label circulates for hours, so the patient can be reimaged when bleeding recurs, which no contrast study allows. It comes after CT angiography and not before it because the answer it returns is different in kind: a bleeding territory on a cine, not a feeding vessel and not a lesion. It localises for the endoscopist, the interventional radiologist or the surgeon, and it is the wrong test in massive haemorrhage, where a ninety-minute acquisition is time the patient does not have.
Problem solving	CT Enterography CT Enterography — Neutral Oral with Enteric-Phase Acquisition	For obscure, recurrent bleeding that has stopped, the target is no longer extravasated contrast but the lesion itself — an angioectasia, a small bowel tumour, a Meckel diverticulum. Neutral luminal distension with an enteric-phase acquisition is what makes an enhancing mucosal lesion visible against a distended, low-attenuation lumen.

A negative study during a quiescent interval does not exclude a source; it dates the bleeding rather than excluding it.
The three studies are ordered by what they detect, not by preference: CT angiography for brisk bleeding now and a feeding vessel, red-cell scintigraphy for slower or intermittent bleeding and for the patient still bleeding after a negative angiogram, enterography for the lesion once the bleeding has stopped.

Pitfalls

Vetting this as a standard portal-venous CT abdomen and pelvis. Without the unenhanced and arterial acquisitions, extravasation cannot be confirmed as new or attributed to a feeding vessel.
Booking the scan for a later list. The sensitivity of the study is a function of whether blood is leaving the vessel at the moment of the acquisition.
Permitting oral contrast, which is the single most destructive protocol error here.
Reporting a negative study as excluding a bleeding source rather than as a negative study during that interval.
Forgetting that bleeding from a variceal source is a different pathway, in which portal venous anatomy and portosystemic collaterals matter more than arterial extravasation.
Stopping at a negative CT angiogram in a patient who is still bleeding. The negative result is evidence about rate, not about the presence of a source, and a bleed too slow for the arterial acquisition is precisely what the labelled red-cell study is sensitive to.
Booking a red-cell scan for a haemodynamically unstable patient. The acquisition takes an hour and a half and localises a territory rather than treating anything; that patient needs CT angiography and interventional radiology.
Reading a delayed static red-cell image without the intervening cine. Bowel is mobile and activity migrates, so a focus labelled "right colon" on a single late frame is the classic mechanism of mislocalisation and of the wrong hemicolectomy.
Accepting a CT angiogram ahead of endoscopy in a stable patient with haematemesis or melaena. Upper endoscopy within 24 hours is both the diagnostic and the therapeutic step, and imaging that displaces it delays treatment rather than adding to it.
Requesting a CT angiogram for occult bleeding — iron-deficiency anaemia with no overt bleeding. There is nothing to extravasate between acquisitions, so the study is negative by construction and the negative result is routinely misread as excluding a source.

Priors — what to pull first

Check whether contrast has been given in the preceding days: residual contrast in bowel or excreted into the renal tract can be mistaken for fresh extravasation, and the unenhanced series is the safeguard.
Review the endoscopy report first — a clipped but re-bleeding lesion tells the interventional radiologist where to look.

What makes a good request

Endoscopy leads this pathway, and radiology joins it rather than opening it. Haematemesis or melaena is resuscitated and then endoscoped — within 24 hours of presentation for a non-variceal upper bleed, and after haemodynamic resuscitation rather than before it in the unstable patient — because upper endoscopy both finds the lesion and treats it, which no scan does. CT angiography earns its place when endoscopy has failed, has been non-diagnostic, cannot be done, or when the bleeding is too brisk for a view; a CT angiogram requested before any endoscopic attempt in a stable patient with haematemesis is the wrong first move.
Timing is the study. Bleeding is intermittent, so the scan should be performed while the patient is actively bleeding rather than booked for the following morning.
Say whether the bleeding is overt and happening now or occult and historical. They are two different cards in practice: extravasation is only detectable during active haemorrhage, and a patient with iron-deficiency anaemia and no overt bleeding needs the lesion found, not the leak caught.
State whether interventional radiology is being considered, because a positive study is usually a prelude to embolisation and the report needs to describe the feeding vessel.
Say what has already been done, and specifically whether a CT angiogram has already been negative. A patient who is still bleeding after a negative angiogram has not had the pathway exhausted — they have had the test that is insensitive to slow bleeding, and labelled red-cell scintigraphy is the next study rather than a repeat of the same acquisition.
Say whether the bleeding is continuous or intermittent. Intermittent bleeding is the specific case in which a labelled pool that persists for hours beats any single-pass contrast study, because the patient can simply be reimaged when it recurs.

Scoring this once it is done

The classification and risk tools this question ends in.

eGFR (CKD-EPI 2021 Creatinine)

How these studies are acquired

Contrast, phases and timing for every study on the pathways above.

CT Mesenteric Angiogram — Biphasic CT Enterography — Neutral Oral with Enteric-Phase Acquisition MR Enterography — small bowel Labelled red-cell scintigraphy — dynamic abdominal acquisition

Confirm locally

CT Mesenteric Angiogram — Biphasic: timings are typical — confirm against local protocol.
Timings, contrast volumes and rates above are typical published values. Your department's protocol, scanner and patient population decide the actual numbers.
Safety thresholds and premedication policy follow local policy where it differs from the cited guidance.

References

Implemented from the cited published sources. Educational and workflow support only; confirm against current guidelines and local policy before clinical use.