Capsule Endoscopy or Angiography in Patients With Acute Overt Obscure Gastrointestinal Bleeding

A Prospective Randomized Study With Long-Term Follow-Up

Wai K Leung MD; FACG; Simon S M Ho MBBS; Bing-Yee Suen BN; Larry H Lai MBChB; Simon Yu MD; Enders K W Ng MD; Simon S M Ng MD; Philip W Y Chiu MD; Joseph J Y Sung MD; PhD; Francis K L Chan MD; James Y W Lau MD


Am J Gastroenterol. 2012;107(9):1370-1376. 

In This Article



During the 30-month study period, 91 acutely bleeding patients with negative upper and lower gastrointestinal endoscopy were assessed (Figure 1). Of these, 11 patients refused consent and 18 patients were excluded for reasons as listed in Figure 1. Two patients remained in shock despite resuscitation and underwent emergency surgery for control of bleeding. In total, 60 patients with acute OGIB were recruited (30 assigned to angiography and 30 to CE). All patients had satisfactory bowel preparation during colonoscopic examination. There were 36 (60%) men and their mean age was 56 (s.d.=20) years. Two patients refused to undergo angiography after randomization and all patients assigned to CE had CE performed. There was no significant difference in the baseline characteristics between the two groups (Table 1).

Diagnostic Yield

With our predefined criteria, CE was positive in 16 patients (53.3%; 95% confidence interval (CI), 36.1–69.8%) and angiography was positive in 6 patients (20%; 95% CI, 9.5–37.3%). The diagnostic yield of CE was significantly higher than angiography (difference=33.3%, 95% CI): 8.9–52.8%; P=0.016).

Findings of CE included three patients with small-bowel tumors (Figure 2) and four patients with active bleeding in small bowel distal to duodenum. Significant small-bowel ulcers and erosions were detected in another six patients. Three patients were noted to have active bleeding lesions in stomach that were not identified by upper endoscopy. For the four patients with active bleeding on capsule endoscopy, subsequent investigations found small-bowel angiodysplasia in two and small-bowel ulcer in one. The remaining patient had no identifiable bleeding source despite further enteroscopy, CT enteroclysis, and Meckel's scan. All capsule endoscopes were spontaneously excreted. If the four patients with active bleeding on CE were regarded as negative finding, the diagnostic yield of CE was 40% (95% CI; 24.6–57.7%) and there was no significant difference in the diagnostic yield between the CE and the angiography groups (P=0.16).

Figure 2.

A bleeding patient with small-bowel gastrointestinal stromal tumor as detected by capsule endoscopy (left) and subsequent computed tomography (CT) scan (right). Red arrows indicate the location of the tumor on capsule endoscopy and CT scan.

In the angiography group, positive findings include one Meckel's diverticulum, one jejunal hypervascular tumor, and four patients with angiodysplasia in small bowel and colon (Figure 3). None of the patients developed complication after angiography.

Figure 3.

Two patients with small-bowel pathology identified by angiography. The left panel shows a hypervascular tumor in jejunum (red arrow) that was subsequently confirmed to be a gastrointestinal stromal tumor. The right panel shows the persistent vitellointestinal artery with contrast extravasation in distal ileum (red arrow) compatible with a bleeding Meckel's diverticulum.

Further Interventions

The details of all subsequent endoscopic and surgical interventions on follow-up period are listed in Table 2. There was no angiographic intervention performed. Four (14.3%) patients in the angiography group developed continuous bleeding during the same hospitalization and were crossed over to CE. Capsule endoscopy identified the bleeding lesions in three patients (two small-bowel diverticular bleeding and one proximal colonic diverticular bleeding). The remaining patient refused further investigation as there was no further rebleeding. No patient in the CE group was crossed over to angiography. Five patients (two in angiography and three in CE groups) underwent surgery for resection of small-bowel tumors or Meckel's diverticulum. Otherwise, there was no significant difference in the numbers of enteroscopy between the two groups.

Long-term Outcomes

The mean follow-up of the study patients was 48.5 (s.d. 20.9) months. In total, 15 (25%) patients developed rebleeding (10 in the CE group and 5 in angiography group). The cumulative rebleeding rate in the CE group and angiography group was 16.7% (95% CI, 7.3–33.6%) and 33.3% (95% CI, 19.2–51.2%), respectively (difference 16.7%, 95% CI, −5.3 to 36.8%; P=0.23). As shown in Figure 4, the cumulative risk of rebleeding tended to be higher in the angiography group than CE group, but the difference did not reach statistical significance (P=0.10, log rank).

There was no significant difference on subsequent hospitalization and transfusion rates between the two groups ( Table 2). Eight (13.3%) patients died within the follow-up period (four in each group) and all deaths were not related to bleeding. Three deaths occurred within the first year of randomization (one in the capsule group and two in the angiography group).Three patients in the CE arm died of stroke and one died of cardiac event. In the angiography group, two patients died of cardiac event, one died of infected foot gangrene, and the other died of aspiration pneumonia.

Seven patients with positive findings in the initial assigned investigation rebled (four in the angiography and three in the CE group). Table 3 summarizes the causes and timing of rebleeding. Overall, four patients in the angiography group rebled from small-bowel or colonic diverticulum and two rebled from benign small-bowel lesions. Two patients with negative angiography had delayed rebleeding from unidentified source. Most patients in the angiography group rebled within the first 3 months. One patient with positive CE had rebleeding because of delay in confirming the diagnosis of small-bowel gastrointestinal stromal tumor. He underwent single-balloon enteroscopy, which failed to identify the lesion, and a subsequent double-balloon enteroscopy during his rebleeding confirmed the small-bowel gastrointestinal stromal tumor. Among the four patients with fresh blood on CE, one of them rebled at 8 months after randomization. However, the bleeding pathology remained undetermined in this patient after extensive investigations as mentioned in the previous session. Two patients with negative CE rebled from proximal small-bowel lesions that were within reach of upper endoscopy.