Assessment of multi-modality evaluations of obscure gastrointestinal bleeding

AIM To determine the frequency of bleeding source detection in patients with obscure gastrointestinal bleeding (OGIB) who underwent double balloon enteroscopy (DBE) after pre-procedure imaging [multiphase computed tomography enterography (MPCTE), video capsule endoscopy (VCE), or both] and assess the impact of imaging on DBE diagnostic yield. METHODS Retrospective cohort study using a prospectively maintained database of all adult patients presenting with OGIB who underwent DBE from September 1st, 2002 to June 30th, 2013 at a single tertiary center. RESULTS Four hundred and ninety five patients (52% females; median age 68 years) underwent DBE for OGIB. AVCE and/or MPCTE performed within 1 year prior to DBE (in 441 patients) increased the diagnostic yield of DBE (67.1% with preceding imaging vs 59.5% without). Using DBE as the gold standard, VCE and MPCTE had a diagnostic yield of 72.7% and 32.5% respectively. There were no increased odds of finding a bleeding site at DBE compared to VCE (OR = 1.3, P = 0.150). There were increased odds of finding a bleeding site at DBE compared to MPCTE (OR = 5.9, P < 0.001). In inpatients with overt OGIB, diagnostic yield of DBE was not affected by preceding imaging. CONCLUSION DBE is a safe and well-tolerated procedure for the diagnosis and treatment of OGIB, with a diagnostic yield that may be increased after obtaining a preceding VCE or MPCTE. However, inpatients with active ongoing bleeding may benefit from proceeding directly to antegrade DBE.


INTRODUCTION
Obscure gastrointestinal bleeding (OGIB) is defined as bleeding from the gastrointestinal (GI) tract that persists or recurs without an obvious etiology after negative upper endoscopy and colonoscopy [1,2] . OGIB is further categorized into "obscure overt" or "obscure occult" bleeding based on presence or absence of evident bleeding and accounts for approximately 5% of all GI bleeding [3] . Though lesions may be missed in the esophagus, stomach, and colon, the etiology of OGIB is secondary to small bowel pathology in up to 75% of cases [4][5][6] leading some experts to recommend that this term be replaced by the term 'small bowel bleeding' [7] . The evaluation of OGIB frequently requires significant utilization of resources and results in patient frustration due to lack of definitive findings and clinical improvement in many cases [8] .
The current American Society of Gastrointestinal Endoscopy guidelines recommend a variety of diagnostic options when evaluating OGIB, with slight differences between the "overt" and "occult" GI bleeding algorithms [9] . At many referral centers, multiphase computed tomography enterography (MPCTE) and/or video capsule endoscopy (VCE) are performed after a negative routine endoscopic exam but prior to double balloon enteroscopy (DBE) as these diagnostic studies are less invasive and may direct DBE-guided therapies [10,11] . MPCTE allows for evaluation of dynamic changes in abnormal enhancement patterns and compares findings across phases in 2-dimensional and 3-dimensional images [12] . Images are evaluated in arterial, enteric and delayed phases allowing for evaluation of ongoing or recent bleeding. In contrast, VCE requires ingestion of a small pill-size camera that provides endoluminal photographs of the entire GI tract for evaluation of small bowel mucosal lesions. These technologies are generally considered complimentary as each can provide different but vital information in the evaluation of OGIB. Though generally performed before DBE, there is a paucity of data regarding how often these tests alter subsequent diagnostic evaluation or treatment.
In this study, we aimed to determine the frequency of bleeding source detection in patients with OGIB who underwent DBE (antegrade/retrograde) after preprocedure imaging (i.e. MPCTE, VCE, or both) and to assess the impact of imaging on DBE diagnostic yield. We also aimed to assess the agreement between findings of the pre-procedure imaging and the DBE itself.

MATERIALS AND METHODS
A retrospective cohort study was conducted following the approval of the Institutional Review Board of Mayo Clinic-Rochester (IRB No. 13-002000 and No. 14-009997). Medical records were reviewed of all adult patients presenting with OGIB who underwent a DBE (antegrade/retrograde) from September 1 st , 2002 to June 30 th , 2013 using a prospectively maintained DBE database. Patients who underwent DBE for indications other than OGIB (i.e. enteral feeding tube placement, failed colonoscopy, evaluation for hereditary polyposis syndromes, small bowel mass, strictures, etc.) were excluded. The electronic medical record was utilized to obtain demographic, endoscopic, radiologic, and clinical outcomes data. At our institution, single balloon enteroscopy is not utilized for assessment of OGIB, and hence this procedure was not included in our study.
Demographic features including age at the time of DBE procedure, gender, and gastrointestinal surgeries prior to DBE were recorded. The total number of blood transfusions up to 30 d prior to the date of DBE procedure and the use of anticoagulant/antiplatelet agents at the time of procedure were collected. Details of VCE and MPCTE performed prior to the DBE were also collected. Only VCE and MPCTE performed within 1 year prior to the DBE procedure were included. All VCE were performed using Pillcam or Pillcam2 (Given Imaging, Yoqneam, Israel). When VCE was performed prior to DBE the date of the procedure, positive and negative findings, and time from VCE to DBE were noted. Positive VCE findings were categorized as (1) arteriovenous malformation (AVM); (2) red spot; (3) frank blood; (4) polyp; (5) ulcer; or (6) other [13] . Similarly, when MPCTE was performed prior to DBE, positive or negative findings and time from MPCTE to DBE were recorded. Positive MPCTE findings were categorized as (1) vascular malformations; (2) blood; (3) polyp/tumor; (4) ulcer; and (5) other [14,15] . Crosssectional imaging findings were abstracted from the final radiologic report.
DBE procedural details included the approach (antegrade vs retrograde), type of obscure bleeding (overt vs occult), and hospital admission status (inpatient vs outpatient). The OGIB was defined as "overt" when the clinician's note reported it to be overt or when there was clinically-evident bleeding including melena or hematochezia reported in the medical records. OGIB was defined as "occult" when the clinician's review reported it to be occult, or when iron deficiency anemia or positive stool testing for blood loss were the sole indication for DBE. We documented whether total enteroscopy was achieved, defined as complete evaluation of the small bowel using either a single approach or combined antegrade-retrograde approach. Findings from DBE were classified into (1) vascular lesions (angioectasias/AVMs, Dieulafoy's lesion, or ectopic varices); (2) mucosal lesions [erythema, erosions, ulcers, inflammation]; or (3) tumor/polyp [16] . If none of the above findings were seen, then the DBE was reported as negative. Therapies performed including argon plasma coagulation (APC), biopsy, hemostatic clip placement, bipolar cauterization, polypectomy and stricture dilation were also recorded. Adverse events including bleeding within 7 d of the procedure, perforation, pancreatitis and re-bleeding within 1 year of the procedure were recorded. Any repeat DBE performed within 1 year of the index DBE was documented.

Statistical analysis
The statistical methods of this study were reviewed by Joseph Larson and Felicity Enders, PhD from the Mayo Clinic Division of Health Sciences Reserch. Continuous measures were summarized using medians and ranges while categorical measures were summarized using counts and percentages. Differences among two groups were assessed using the Kruskall-Wallis test and Chisquare or Fisher's exact test for continuous or categorical measures, respectively.
To evaluate the predictive ability of VCE and MPCTE to identify bleeding sites, DBE was treated as the gold standard and the sensitivity, specificity, diagnosis yield, and accuracy were calculated among patients with VCE and MPCTE within one year of DBE. Ninety-five percent confidence intervals for each of these measures were also determined. This analysis was repeated among the following subgroups; antegrade and retrograde approach, inpatient and outpatient procedure, overt and occult bleeding.
Because the same patients underwent VCE or MPCTE and DBE, to assess the findings from the procedures, matched logistic regression performed with the finding treated as the outcome and the DBE test treated as the predictor. Odds ratios and 95% confident intervals along with P values are presented for these tests.
All analyses used an significance level of 5% and were performed using the SAS (v9.3, SAS Institute Inc., Cary, NC, United States).     In order to compare findings on DBE to VCE and MPCTE, matched odds ratios were examined. There were no increased odds of finding a bleeding site at DBE compared to VCE (OR = 1.3, 95%CI: 0.9-1.7, P = 0.150). There were increased odds of finding a bleeding site at DBE compared to MPCTE (OR = 5.9, 95%CI: 3.5-9.7, P < 0.001).

DISCUSSION
Small bowel bleeding is the commonest cause of OGIB, seen in 75% of cases [17] . Identifying the site of bleeding and its therapy remain challenging due to this anatomic location. DBE is an effective way to address these challenges but is costly and not readily available at all centers. Our study aimed to look at the diagnostic yield of DBE and of preceding VCE and MPCTE. This would allow us to analyze the need for imaging prior to DBE. In our large single center cohort of 495 patients with OGIB, the yield of DBE without preceding VCE or MPCTE was 59.4%, and with preceding imaging was 67.5%. Thus, although the diagnostic yield of DBE is higher when pre-DBE imaging is positive, a source lesion is frequently identified when pre-DBE imaging is negative or not performed.
Using direct visualization by DBE as the gold standard, VCE had a diagnostic yield of 72.7% but a relatively low specificity of 28.8%. This is similar to prior studies [18] . The commonest findings at VCE with negative DBE were AVMs and blood; it is possible that these abnormalities had subsided by the time of the DBE since the time interval between the tests in our study could be up to 1 year [11,19] . This would be characteristic of AVMs which often bleed intermittently, and could artificially increase the apparent false positive rate.
A preceding MPCTE was done in fewer patients compared to VCE and had a lower diagnostic yield of 32.5%. However, the specificity was higher at 71.6%, also similar to previous studies [14,20,21] . The vascular lesions seen in nearly half the patients with positive MPCTE and negative DBE were likely deep in the bowel wall and hence not seen endoscopically.
Antegrade DBEs overall had higher diagnostic yields than retrograde DBEs (72.51% vs 50.91%, P < 0.001. This has been shown in one other smaller series [22] . Thus, the overall diagnostic yield of antegrade DBE is superior to CTE and roughly equivalent to VCE in the evaluation of OGIB. This is an important finding because it suggests that almost all patients should undergo antegrade DBE before retrograde, unless otherwise dictated by abnormal MPCTE suggesting ileal tumors or polyps. The diagnostic yields of DBE for inpatients vs outpatients were similar in our data but the highest sensitivity of VCE using DBE as gold standard was in inpatients (84.9%). This group also showed the highest specificity (45.0%). Interestingly the incremental diagnostic yield of DBE of all patients with negative preceding VCE and MPCTE was 66% (35/53 patients). Thus, this raises the question of whether an appropriate strategy might be to directly proceed to antegrade DBE in inpatients with evidence of ongoing bleeding if DBE is available. This is also reflected by the matched odds ratios comparing VCE and MPCTE to DBE where there were no increased odds of finding a bleeding site at DBE compared to VCE and increased odds at DBE compared to MPCTE. In our data, none of the tests had a significantly higher yield in patients with overt bleeding compared to occult bleeding, which is unlike prior studies [23][24][25] In conclusion, our data suggest that DBE is a generally safe and well tolerated procedure for the diagnosis and treatment of OGIB, with a diagnostic yield that may be increased after obtaining a preceding VCE or MPCTE. However, inpatients with active ongoing bleeding may benefit from proceeding directly to antegrade DBE, which has the benefits of improved diagnostic yield in these patients, ability to intervene therapeutically, and avoidance of an additional diagnostic test. A prospective evaluation and cost-effectiveness analysis of this clinical algorithm would be warranted.

Background
The etiology of obscure gastrointestinal bleeding (OGIB) is secondary to small bowel pathology in up to 75% of cases. The authors sought to determine the frequency of bleeding source detection in patients with OGIB who underwent double balloon enteroscopy (DBE) after pre-procedure imaging (multiphase computed tomography enterography, video capsule endoscopy, or both) and assess the impact of imaging on DBE diagnostic yield.

Research frontiers
Diagnostic yields of DBE, computed tomography enterography and video capsule enteroscopy in obscure gastrointestinal (GI) bleeding.

Innovations and breakthroughs
This is the one of the largest cohort of patients with occult GI bleeding undergoing DBE for occult GI bleeding. A large proportion of patients also had preceding imaging, allowing for comparison of the various techniques.

Applications
Inpatients with active ongoing bleeding may benefit from proceeding directly to antegrade DBE without preceding testing, which has the benefits of improved diagnostic yield in these patients, ability to intervene therapeutically, and avoidance of an additional diagnostic test.

Terminology
Obscure GI bleeding -GI bleeding where etiology is not in the esophagus, stomach, or colon.