CT angiograms for lower GI bleeding: the experience of a large UK teaching hospital

Abstract

Introduction

The clinical presentation of lower gastrointestinal bleeding (LGIB) is variable in severity, cause and potential investigations. The British Society of Gastroenterology recently published LGIB guidelines, recommending CT angiography (CT-A) for haemodynamically unstable patients, defined by shock index (SI) greater than 1. The aim of this study was to assess the use and role of CT-A in diagnosing LGIB, by assessing the pickup rate of active LGIB defined by contrast extravasation or ‘blush’ and to determine any association between positive CT-A with various patient and clinical characteristics.

Methods

A retrospective analysis was carried out of 4 years of LGIB admissions. Demographics, inpatient observations and use of blood products were acquired. Vital signs nearest the time of CT-A plus abnormal vital signs preceding imaging were used to calculate SI, Age SI, National Early Warning Score 2 (NEWS2) and Standardised Early Warning Score (SEWS). A consultant gastrointestinal radiologist further reviewed all consultant-reported scans.

Results

In total, 930 patients were admitted with LGIB. Median age was 71 years and 51% were male; 179 (19.2%) patients received red blood cell transfusion and 93 patients (10%) underwent CT-A, who were older and were likely to be hypotensive and receive red cell transfusions. Following exclusions, 92 CT-As were included in the analysis. Nine (9.8%) were positive. Univariate analysis showed no association between positive CT-A and any scoring system. A multivariate analysis, including age and gender, showed association between both NEWS2 and SEWS scores with positive CT-A.

Conclusion

In our analysis of the typical LGIB population, CT-A has shown relatively low pick up rate of active bleeding. CT-A clearly has a role in the investigation of LGIB, but selection remains challenging.

Introduction

Gastrointestinal bleeding is traditionally divided into lower (LGIB) and upper gastrointestinal bleeding (UGIB). LGIB is defined as bleeding that originates distal to the ligament of Trietz. UGIB, in contrast to LGIB, is amenable to diagnosis and treatment via oesophago-gastro-duodenoscopy (OGD) and is associated with higher mortality rates.¹ Thus, LGIB requires investigations and treatment beyond OGD, which include abdominal CT scanning, CT angiography (CT-A), mesenteric angiography, flexible sigmoidoscopy or colonoscopy and occasionally surgery.² The clinical presentation of LGIB varies by onset and severity and has a plethora of potential causes.³

In the UK, the British Society of Gastroenterology (BSG) has recently published guidelines on the diagnosis and management of LGIB.⁴ It is recommended that CT-A is performed in patients with LGIB who are haemodynamically unstable in order to localise the site of bleeding prior to therapy. CT-A is an imaging modality that uses intravenous contrast in three phases (non-contrast, arterial, venous) to assess for haemorrhage via contrast extravasation from the intravenous space.⁵ A positive CT-A result is where contrast extravasation or ‘blush’ is visible on CT imaging, thus identifying the anatomical location of haemorrhage and guiding further treatment such as embolisation, endoscopy or surgery.

CT-A has a higher diagnostic yield in patients who are haemodynamically unstable,^6,7 leading the recent BSG guideline to recommend use of the shock index (SI, heart rate/systolic blood pressure) to select patients for CT-A.⁴ The guideline cites identification of the source of bleeding in 49.7% of cases undergoing investigation with CT-A.

The Queen Elizabeth Hospital Birmingham, part of the University Hospitals Birmingham NHS Foundation Trust, is a major trauma centre and has 24-hour access to CT-A and interventional radiology. Having had ready access to CT-A for many years, we wished to determine our local success rates in identifying the source of bleeding using this modality. The aim of this study was to assess the use and accuracy of CT-A in the diagnosis of lower GI bleeding and to determine any association between positive CT-A results with age, vital signs and derivative metrics.

Methods

A retrospective analysis of all patients admitted with lower GI bleeding over a period of 4 years, from January 2015 to January 2019, was performed. All admitted patients require an admission diagnosis selected from a pick list. Patients were thus identified by interrogating the electronic admissions database for the admission diagnosis of LGIB. Basic demographic data were retrieved and our electronic Prescribing Information and Communication System (PICS) was searched to obtain vital signs on attendance and throughout admission. Vital signs closest to the time of imaging as well as abnormal vital signs in the period of time preceding imaging were used to calculate the SI and Age SI. SI was calculated using heart rate divided by systolic blood pressure. Age SI was calculated by multiplying the shock index by the patient age.

We calculated the National Early Warning Score 2 (NEWS2) and Standardised Early Warning Score (SEWS) when all the required information was available.⁸ CT angiograms were reported by a consultant radiologist during admission. We subsequently arranged for all CT angiograms to be reported independently by a second consultant radiologist with a specialist interest in gastrointestinal radiology.

Demographic, observational and derived variables were tested for association with a positive outcome on CT-A by both univariate and multivariate (binomial logistic regression) analyses. Differences between continuous and ordinal variables between groups were tested for significance using the Mann–Whitney test, and differences in categorical values, by the Pearson’s chi-square test. All computations and analysis was performed using R (R Foundation for Statistical Computing, Austria). p-values <0.05 were deemed statistically significant.

CT-A was performed on Siemens 65 slices and Siemens 128 slices. Spiral acquisition was used with a slice thickness of 0.6mm. Reconstruction slice thickness was 1mm. The contrast protocol was an 80ml volume given at a rate of 4ml/s. Saline flush 18ml was given before contrast, followed by contrast injection, then 25ml saline chaser. The scanning protocol involved plain CT abdomen and pelvis, followed by further scans at 40 and 90 second intervals following the start of contrast administration.

This study received institutional approval and was registered as an audit (audit identifier: CARMS-15047).

Results

In total, 930 patients were identified who were admitted with a primary diagnosis of LGIB. For all LGIB patients, median age was 71 years (IQR 54–83), 51% of whom were male. Overall, 179 (19.2%) patients received red cell transfusion, and 143 (15.4%) patients received tranexamic acid. Ninety-eight (10.5%) patients were regularly taking antiplatelet medications; 29 (3.1%) patients were regularly taking coumarin anticoagulants; and 18 (1.9%) novel oral anticoagulants. Overall, 115 (12.4%) patients experienced at least one episode of hypotension during admission, defined as a systolic blood pressure <90mmHg. The median lowest recorded haemoglobin during admission was 103g/l (83–129).

Ninety-three patients (10%) underwent CT-A. Comparisons between those patients undergoing CT-A and those who did not are listed in Table 1. The CT-A group were older (median age 74 vs 70), more likely to be male (70% vs 49.1%), had lower haemoglobin (median 80.0g/l vs 108g/l), more likely to require red cell transfusion (72% vs 13.4%) and were more likely to have an episode of hypotension (43.0% vs 9.4%).

Table 1 .

Comparison of demographic and admission variables, grouped by whether patients underwent CT-A

Variable	All patients n=930	CT-A group n=93	No CT-A group n=837	p-value
Age, median (IQR)	71.0 (54.0–83.0)	74.0 (65.5–86.0)	70.0 (52.5–84.0)	0.018*
Gender male – n (%)	474 (51.0)	63 (70.0)	411 (49.1)	0.002*
Antiplatelets – n (%)	98 (10.5)	15 (16.1)	83 (9.9)	0.094
Anticoagulants – n (%)	47 (5.32)	8 (8.60)	39 (4.65)	0.162
Requiring red cell transfusion – n (%)	179 (19.2)	67 (72%)	112 (13.4)	<0.001*
Median haemoglobin g/l (IQR)	103 (83.0–129)	80.0 (67.0–91.0)	108 (87–132)	<0.001*
Episode of hypotension – n (%)	115 (12.4)	36 (43.0)	79 (9.4)	<0.001*

All data given to three significant figures.

CT-A = CT angiography; g/l = grams/litre; IQR = inter-quartile range. Hypotension defined as systolic blood pressure <90mmHg.

*Significant p-value.

CT-A analysis

Ninety-three (10%) individual patients underwent CT-A, with a total of 107 CT angiograms performed, including 13 patients undergoing repeat CT-A on admission. No repeat CT-A was positive when the first was negative.

Following our second expert radiological review, 13 scans were excluded. The most common reason for exclusion was presence of oral contrast in the GI tract (n=6). A further two CT-As had no observational data and were therefore not included in the SI calculations. Seventy-eight patients undergoing 92 individual CT angiograms were included in our SI analysis.

Of the 92 CT angiograms included in the analysis, 9 scans were positive (9.8%). One patient had two positive scans. Although without evidence of active bleeding, 17 CT angiograms did show evidence of colitis or diverticulitis.

CT-A outcome and patient characteristics

A univariate analysis of variables and CT-A outcome was completed. Comparisons between those patients with positive CT angiograms and those with negative CT angiograms are included in Table 2. Although the positive CT-A group were older (mean 84 years vs 73 years), more likely to be male (89% vs 69%), had a greater highest heart rate (mean 106 vs 95.5) and lesser lowest blood pressure (mean 83 vs 93.5), none of these differences were statistically significant.

Table 2 .

Results of univariate analysis comparing demographic, vital signs and derived variables, grouped by CT-A outcome (positive/negative)

Variable	CT-A positive n=9	CT-A negative n=83	p-value
Age	83 (65–90)	73 (67–81)	0.511
Gender male – n (%)	8 (89)	57 (69)	0.199
HR at time of CT-A	89 (85–103)	82 (72.5–98.5)	0.378
Sys BP at time of CT-A	111 (99.0–115)	117 (105–133)	0.152
SI at time of CT-A	0.860 (0.760–0.970)	0.700 (0.570–0.835)	0.132
Age SI at time of CT-A	64.5 (20.8)	54.2 (17.8)	0.112
NEWS2 at time of CT-A	6.00 (5.00–10.00)	5.00 (3.00–6.00)	0.102
SEWS at time of CT-A	4.00 (2.00–5.00)	3.00 (2.00–3.00)	0.133
HR highest	106 (102–120)	95.5 (78.8–107)	0.289
Sys BP lowest	83 (81–94)	93.5 (87.0–104)	0.429
SI highest	1.00 (0.860–1.14)	0.80 (0.625–0.985)	0.247
Age SI highest	80.7 (49.1–96.0)	55.7 (46.1–75.4)	0.222
NEWS2 highest	5.00 (3.50–6.50)	4.00 (2.00–7.00)	0.102
SEWS highest	3.00 (1.75–4.25)	1.00 (1.00–3.00)	0.133

All data quoted as median (IQR) to three significant figures.

Age SI = age×shock index; CT-A = CT angiography; HR = heart rate; IQR = inter-quartile range; NEWS2 = National Early Warning Score 2; SEWS = Standardised Early Warning Score; Sys BP = systolic blood pressure; SI = shock index.

However, by multivariate analysis both NEWS2 and SEWS score were independently associated with a positive CT-A outcome when a binomial logistic regression model was constructed using age, gender and NEWS2/SEWS as covariates (see Table 3).

Table 3 .

Results of binomial logistic regression models using CT-A outcome (positive/negative) as the outcome variable

Variable	Odds ratio (95% CI)	p-value
Age	1.02 (0.960–1.09)	0.482
Gender	4.59 (0.521–40.5)	0.17
NEWS2	1.30 (1.04–1.63)	0.023*
Age	1.03 (0.966–1.09)	0.375
Gender	4.27 (0.492–37.1)	0.188
SEWS	1.51 (1.01–2.25)	0.043*

The model was run using age, gender and either NEWS2 or SEWS as the covariates.

All data given to three significant figures.

CI = confidence interval; NEWS2 = National Early Warning Score 2; SEWS = Standardised Early Warning Score.

*Significant p-value.

Post-scan outcomes

Of the eight patients who had positive CT angiograms, six went on to have angiography. Three patients had positive contrast extravasation identified and four underwent embolisation. Of the 70 patients who had negative CT angiograms, 3 went on to angiography, with 1 patient having positive contrast extravasation and all 3 patients undergoing embolisation. See Appendix 1 for more details on these three cases.

In total nine patients proceeded on to angiography and seven patients undergoing embolisation. No patients underwent angiography/embolisation without a prior CT-A. One patient proceeded to laparotomy. This patient had a small bowel resection for a small bowel lymphoma, identified as a small bowel mass on CT-A.

Forty-eight patients (51.6%) subsequently underwent lower GI endoscopy, either sigmoidoscopy, colonoscopy or both; 36 patients (38.7%) underwent upper GI endoscopy. Outcomes from upper and lower GI endoscopy are found in Table 4. Patients in the CT-A group were more likely to undergo both lower and upper GI endoscopy in comparison to the non-CT-A group, where 360 underwent lower GI endoscopy (43%) and 72 (8.6%) underwent upper GI endoscopy. Two patients were found to have probable malignant tumours on lower GI endoscopy and seven were found to have colitis.

Table 4 .

Outcomes from lower and upper GI endoscopy in the CT-A group

	Upper GI endoscopy	Lower GI endoscopy
		Sigmoidoscopy	Colonoscopy
Total	36	34	19
Completed as inpatient	36 (100%)	29 (85.3%)	3 (15.8%)
Completed as outpatient	0	5 (14.7%)	13 (68.4%)
Median time to endoscopy	1.5 days	3.5 days	22 days
Active bleeding identified	2 (6.06%)	0	0
Therapy/intervention performed	7 (19.4%)	1 (2.9%)	2 (10.5%)

Discussion

LGIB is a common presenting complaint, making up 3% of referrals to general surgery.⁹ In contrast to UGIB, there have been few guidelines detailing the investigation and management of LGIB. This may be because of a more limited evidence base for LGIB and available mortality data describing LGIB as a self-limiting condition that rarely results in harm.¹ A recent UK audit of LGIB has however demonstrated that LGIB is associated with significant mortality: 3.4% and 17.9% when occurring in the inpatient population.¹⁰ Over a quarter of patients in this audit received a red cell transfusion yet nearly half did not undergo any inpatient investigations. Early investigation and treatment could lead to reduction in blood product transfusion, thus avoiding transfusion-related complications for patients and reducing NHS transfusion costs.¹ While UGIB is within reach of an endoscope, the investigation of LGIB may involve abdominal CT, CT-A, flexible sigmoidoscopy/colonoscopy, mesenteric angiography and red cell scintigraphy.^11–13

BSG guidelines detailing the diagnosis and management of LGIB were published last year and recommended the use of CT-A in haemodynamically unstable patients or patients with a SI>1 after initial resuscitation, or if active bleeding is suspected. The basis for this recommendation is that CT-A provides the fastest and least invasive means to localise blood loss prior to therapy.⁴ The UK audit reported a diagnostic yield of 49.7% from CT-A perhaps underpinning the recommendation for its use in the guidelines. However, only 5.9% of patients in the audit actually underwent CT-A.

In our study we report a positive CT-A rate of 9.8%, significantly lower than the UK audit. We were strict with definitions in that a CT-A was only deemed positive if there was evidence of active bleeding defined by contrast extravasation (blush). The presence of blood or clot on the plain CT indicates acute GI bleeding but is not reliable in specifying the site or cause of bleeding. There is no proof that any clot we see in the GI tract is adherent to the site of bleeding, as blood tends to move along the GI tract. Additionally, occasional colitis or diverticulitis was noted on CT imaging. These findings may have been the cause of LGIB, and some authors may include these findings in the definition of a positive test. In our study, the CT-A positive rate would have increased to 28% if such cases were included. The presence of uncomplicated diverticulosis is a common CT finding¹⁴ and should certainly not define a test as positive. It is not clear how the diagnostic yield of CT-A was defined in the UK audit. Allowing irrelevant CT-A findings to be counted as the likely source of bleeding could explain differences in our CT-A positive rates. We were careful to dual-report CT-As to ensure accuracy of the reports. Nevertheless, our CT-A positive rate is low compared with many other studies, including a meta-analysis, that report a rate of between 14% and 91%.^{6,7,11–13,15–19} Some studies have highly selected patient groups, including those who proceed to surgery,¹⁶ as well as further mesenteric angiography.¹² It is known that diagnostic studies with non-representative patients overestimate the accuracy of diagnostic tests.²⁰ Our study is representative of the population of patients presenting with LGIB to emergency departments and acute surgical units in the UK, and we therefore feel it is important to report that the sensitivity of CT-A for lower GI bleeding in this patient group may not be as high as that reported in previous studies or the recent UK audit. Our study was however retrospective, and there was no specific guideline or algorithm that was followed to decide on which patients should undergo CT-A. Furthermore, patients in our study had lower levels of antiplatelet medication (10.5% vs 23.1%) and coumarin-based anticoagulation (3.1% vs 10.7%) than in the UK National Audit. This may be explained by the retrospective nature of our study, or our study population might comprise a group of patients with milder LGIB.

It is clear from experimental studies that CT-A is incredibly accurate, having the ability to detect bleeding rates of <1ml/min.²¹ The accurate selection for CT-A of patients who are acutely bleeding is therefore the key determiner of the test positive rate. It has been shown that haemodynamic instability,^6,7 as well as prompt scanning,¹⁷ is associated with a positive CT-A. In keeping with this, the BSG guidelines recommend the use of SI to select patients for CT-A.⁴ Prior to the BSG guidelines, we had not followed any clear selection criteria for CT-A, rather selection was based upon a clinical suspicion of ongoing bleeding determined by senior decision makers. CT-A may have been used liberally in our study as it is non-invasive and easy to access and may explain our low test positive rate. Nevertheless, there was definite evidence for selection in our study, as patients undergoing CT-A were significantly more haemodynamically unstable with lower haemoglobin and significantly more likely to receive red cell transfusion than those who did not undergo CT-A.

SI is a simple and easily calculated score that was initially applied to major trauma as a marker of haemorrhagic shock and mortality.²² However, Age SI has been shown to be a better predictor of mortality in the elderly population than SI.²³ The median age of UK patients presenting with LGIB is 74 years,¹⁰ and thus the same concerns arising from use of SI in the elderly would apply to this LGIB cohort. In our study, there was no significant association between CT-A outcome and SI, nor Age SI, but there were only nine positive tests. Even despite this low number of positive test results, we were able to show a significant association between CT-A outcome and scores of illness severity, namely NEWS2 and SEWS scores. The intermittent nature of LGIB means than timing for CT-A is important.¹⁷ We examined vital signs closest to the time of CT-A as well as the most extreme vital signs prior to imaging and only demonstrated an association between CT-A outcome and NEWS2/SEWS scores closest to the time of CT-A. This supports the view that CT-A must be performed promptly when active bleeding is clinically suspected in order to yield a positive result.

There is clearly a role for CT-A in the investigation of LGIB. It can be performed rapidly when available, is non-invasive and relatively operator-independent.¹¹ Some variation in the performance of CT-A can be attributed to the technical parameters of the scan, and so it is important to have consensus and consistency in the radiology department to optimise the performance of the test.^19,24 A negative CT-A outcome may be able to facilitate earlier discharge from hospital.²⁵ CT-A may also identify likely causes of the bleeding such as tumours, colitis and diverticulitis even if no active extravasation of contrast is seen.¹ It could be considered an unnecessary test given that the vast majority of LGIB stops spontaneously and will be subject to later endoscopic evaluation. We feel that while CT-A is a useful test for LGIB for reasons stated above, its diagnostic yield may be overstated in the BSG guidelines, based on the UK audit, where only 5.9% of patients underwent this test. In contrast to the BSG guidelines, the American College of Gastroenterology published a LGIB guideline in 2016 in which CT-A did not feature as a recommended first-line investigation.²⁶ Similarities do exist such as the initial management focus on haemodynamic resuscitation and consideration given to an UGIB presenting as a LGIB in patients with haemodynamic instability. In the American guidelines, colonoscopy was favoured as the investigation of choice owing to its high diagnostic yield and ability to provide endoscopic therapy. A randomised-controlled trial comparing early colonoscopy with angiographic-based treatment showed improved diagnostic yield in the early colonoscopy group (42% vs 22%) although no difference in any patient outcome measures.²⁷ Effective colonoscopy requires availability of an endoscopist as well as bowel preparation; even though this may seem impractical in the emergency situation, it is possible to perform effective colonoscopy after rapid preparation with polyethylene glycol.^27,28 Colonoscopy also featured as the investigation of choice for severe LGIB in the Scottish Intercollegiate Guidelines Network, albeit published in 2008.²⁹ Certain patient groups may be particularly suited to urgent colonoscopy rather than CT-A. Post-polypectomy bleeds are known to be localised to the colon, and patients may already have adequate bowel preparation. CT-A would not add any useful diagnostic information in such patients.¹⁸

Conclusion

Correct identification of patients as having stable or unstable LGIB is complex and challenging. CT-A has recently been recommended as the primary investigation for unstable LGIB,⁴ but a case can be made for urgent colonoscopy, especially in certain patient groups. Further research is needed to determine which patients require urgent CT-A to localise the bleeding source, while providing a high diagnostic yield so as not to waste resources and the opportunity for other therapies.

Ethics approval

This study did not require formal ethical approval and was approved by the Clinical Audit Department, University Hospitals Birmingham NHS Foundation Trust, reference: GHH-5000.