Introduction Gastrointestinal (GI) bleeding cannot always be controlled or identified at endoscopy (OGD), therefore guidelines recommend radiological intervention in an unstable patient. The options are fluoroscopic angiogram (FA) which is time consuming, requires significant expertise, and has significant radiation exposure. Alternatively a CT angiogram (CTA) may offer a sensitive, rapid diagnosis of the source of GI bleeding to allow definitive treatment. Data on the role and diagnostic yield of CTA are lacking. We present the largest international study to assess the impact of CTA in upper GI bleeding.

Method A retrospective analysis of endoscopy and radiology databases was used to identify patients who underwent radiological intervention for GI bleeding at Sheffield Teaching Hospitals between 2003 and 2013. A control group of 757 patients (who did not undergo FA or CTA) from the South Yorkshire GI bleed audit 2011 was used for comparison. Pre-endoscopy Rockall scores, routine haematology and biochemistry results taken prior to endoscopy and inpatient mortality rates were compared.

Results 59 patients (35 male, mean age 69.3) underwent imaging for upper GI bleeding during the study period.

In 49/59 (mean age=68.3, mean Rockall = 3.8, mean Hb = 84.7) a source was initially identified at OGD but with continued bleeding. CTA was performed in 31 with the other 18 having immediate FA. For patients having a CTA first a source was identified in 15 (48%) requiring embolization in 10. 4 patients who did not have an abnormality seen on CTA subsequently required FA and embolization and 2 of these patients died. For the other 10/59 (16.9%) (mean age 74.3 mean Rockall=4, mean Hb=84.2) no source was identified at OGD and they underwent CTA identifying 8 bleeds which proceeded to FA and then embolization in 6. Inpatient mortality in this group was 50% which was significantly higher than 14.3% for those patients with an identified source for upper GI bleed (p = 0.02).

Patients who underwent CTA were older (p = 0.039) and presented with higher median pre-endoscopy Rockall scores (p = 0.003) than controls. Both CTA and ‘direct to FA patients’ presented with lower Hb (p < 0.0001) than controls. There was no significant difference between CTA and FA patients.

Inpatient mortality rates were higher in those who underwent CTA prior to FA (22%) compared to those who went directly to FA (11%) but this was not significant (p = 0.5).

Conclusion Radiological imaging resulted in embolization in 55.9% of patients referred. Patients requiring radiological intervention for GI bleeds have more significant bleeding with lower haemoglobin and higher Rockall scores. Patients with no source of bleeding on OGD have higher inpatient mortality despite radiological intervention.

Disclosure of interest None Declared.