ORIGINAL ARTICLE
Is Endoscopic Therapy Effective for Angioectasia in Obscure Gastrointestinal Bleeding? A Systematic Review of the Literature Joseph Romagnuolo, MD, MSc, FACG, FASGE,* Andrew S. Brock, MD,w and Nathaniel Ranney, MDw
Goal: We aimed to summarize pooled rebleeding rates of angioectasia after therapeutic endoscopy, and compare these to historical control (no intervention) rates. Background: Obscure gastrointestinal bleeding continues to be challenging to diagnose and treat; in America, small bowel angioectasias are the most common cause. Technology advances led to higher diagnostic yield for these lesions; however, therapeutic impact of endoscopy remains unclear. Study: A PubMed search (June 1, 2006 to September 19, 2013) with 2 independent reviews sought articles reporting rebleeding rates of symptomatic angioectasia without therapy (natural history) and after endoscopic treatment. This study list was added to studies in the 2007 American Gastroenterological Association systematic review. Data on number of patients who underwent endoscopic therapy, type of therapy used, number of patients who experienced rebleeding, and follow-up time were extracted. Rebleeding data were pooled and weighted averages were reported with 95% confidence intervals (CI). Results: Twenty-four articles (n = 490 patients) with data on endoscopic therapy for angioectasia and 6 natural history cohorts (n = 130) receiving no therapy for angioectasia were eligible. Of the endoscopic therapy patients, 121 at push enteroscopy and 427 at balloon-assisted enteroscopy; 209/490 (42.7%; 95% CI, 38%-47%) rebled. Of the control (no therapy) patients, 64/130 (49.2%; 95% CI, 40%-58%) rebled. Number needed to treat is estimated at 15 to 16. Conclusions: Rebleeding rate after endoscopic therapy for symptomatic small bowel angioectasia may be comparable to that expected without therapy. Endoscopic therapy may be ineffective; if effective, the needed to treat is estimated to be high. Controlled studies, with intervention-stratified and etiology-stratified outcomes are needed. Key Words: obscure GI bleeding, angioectasia/angiodysplasia, rebleeding, endoscopic therapy, outcomes, natural history, small bowel/intestine
(J Clin Gastroenterol 2015;49:823–830)
Received for publication May 15, 2014; accepted October 9, 2014. From the *Department of Public Health Sciences, Medical University of South Carolina; and wDepartment of Medicine, Division of Gastroenterology and Hepatology, Medical University of South Carolina, Charleston, SC. Presented (by J.R.) in part at ACG 2009 and ACG 2011 at topic forums: “Is endoscopic therapy effective for angioectasia?” and “Endoscopic therapy for angioectasia: are we treating ourselves or the patient?,” respectively. J.R.: concept and design; J.R., A.S.B., N.R.: data collection and analysis; manuscript drafting; critical revision of manuscript. J.R. was funded by an ASGE Senior Investigator Mentoring Award. The remaining authors declare that they have nothing to disclose. Reprints: Joseph Romagnuolo, MD, MSc, FACG, FASGE, Palmetto Health, Columbia Gastroenterology Associates, 2739 Laurel St, Ste 1A, Columbia, SC 29204 (e-mail: [email protected]). Copyright r 2014 Wolters Kluwer Health, Inc. All rights reserved.
J Clin Gastroenterol
A
n angioectasia is a collection of ectatic blood vessels made of a thin wall with or without an endothelial lining, and with or without coexisting arteriovenous communications.1 It is also referred to as arteriovenous malformation, vascular lesion, venous ectasia, angioma, or angiodysplasia. It is most often incidentally found at endoscopy, and 1 g/dL. Retrospective. 133 of 261 had vascular lesions, 129 treated endoscopically (APC), with 31 lost to follow-up. Most (121 of 129) had angioectasia (5 Dieulafoy, 3 Blue Rubber Bleb Nevus syndrome). Rebleeding defined as reappearance of overt bleeding, refractory or iron deficiency anemia, or continued need for transfusion. Retrospective. 5 of 75 had angioectasia, treated endoscopically (APC, epinephrine injection, and/ or clip). Rebleeding defined as hematochezia, melena, or need for blood transfusion. Retrospective. 79 of 147 had angioectasia, treated endoscopically (details not given), with 16 lost to follow-up (8 had lesions “proximal to the ligament of Treitz” which had 38% rebleeding). Rebleeding defined as overt bleeding, hospitalization, need for further intervention, or need for blood transfusion or iron therapy. Retrospective. 10 of 113 had angioectasia, 3 treated endoscopically (APC). Rebleeding defined as overt bleeding or otherwise unexplained hemoglobin drop >2 g/dL compared with baseline. Retrospective. 26 of 39 had angioectasia, treated endoscopically (APC or heater probe). Rebleeding not well defined.
*202/478 = 42.3% if the 4 studies with n < 5 excluded. APC indicates argon plasma coagulation (noncontact coagulation).
smallest (n = 5)39study is excluded] (Table 3, Fig. 1). The control groups did not appear to be comprised of low-risk individuals. In fact, half of the patients in 1 trial38 had lesions prominent enough to be seen on angiography. In addition, most patients enrolled in the 6 studies had a history of recurrent/chronic bleeding rather than a single episode; single episodes were also often hemodynamically significant. The most powerful natural history data are likely that of the randomized trial of hormonal therapy versus placebo, wherein the rebleeding rate without therapy was only 46% at 2 years, despite high-risk presentations.38 The weighted summary rebleeding rate, pooling data from the 6 studies listed, was 64 bleeds in 130 patients or 49.2% (95% CI, 40%-58%), over 1.5 to 2 years.
Rebleeding After Endoscopic Treatment of Small Bowel Angioectasia The rebleeding rates after endoscopic treatment are more difficult to tease out. Most trials have centered on the diagnostic test rather than the treatment, and so, stratified
rebleeding rates (treated vs. untreated, endoscopic vs. other treatment) are difficult to extract. More than half of studies involving enteroscopy do not report rebleeding rates. Many do not report angioectasia-specific outcomes, or describe the therapy for angioectasias. In the papers that do report outcomes after endoscopic therapy for angioectasia patients, the corresponding rebleeding rates were variable from study to study, and from procedure to procedure (Table 2, Fig. 1). With push enteroscopy, the range in rebleeding rates was 0% to 66%, with a weighted mean of 34.9% (6 studies, n = 63). With deep (balloon-assisted) enteroscopy, the range in rebleeding rates was 24% to 67%, with a weighted mean of 43.8% (18 studies, n = 427). The pooled data from these 24 studies included 490 patients, of which 209 patients rebled after endoscopic intervention, for an overall weighted rebleed rate of 42.7% (95% CI, 38%-47%), over 1.5 to 2 years. This estimate has CIs that overlap markedly with those of the natural history pooled rate; it is also higher than that from the surgical resection studies after lesion identification
826 | www.jcge.com Copyright r 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Clin Gastroenterol
Volume 49, Number 10, November/December 2015
Effectiveness of Endoscopy for Angioectasia
TABLE 3. Natural History Studies of Rebleeding of Angioectasia
n
Rebleeding Rate n (%)
Average/median Follow-up (y)
36
17 (47)
3
Lewis et al37
34
19 (56)
1
Junquera et al38
35
16 (46)
2
Saurin et al30
13
6 (46)
1
Park et al39
5
4 (80)
2.6
Koh et al34
7
2 (28)
2 (mean)
64 (49.2)/95% confidence interval, 40%-58%*
1.9
References Richter et
al36
Totals
130
Comments Retrospective. Medically treated. Colonic angioectasias included.* Rebleeding defined as needing transfusion, hospitalization, or surgery. Retrospective. Patients who refused hormonal therapy or prehormonal therapy era. Rebleeding defined as needing transfusion. Placebo arm of randomized trial of hormonal therapy for angioectasia diagnosed by endoscopy and angiography. Rebleeding defined as acute bleed vs. anemia refractory to iron therapy with positive fecal occult blood Retrospective analysis of previous prospective study comparing push endoscopy and capsule endoscopy. Rebleeding defined as presence of persistent anemia, transfusion requirement, or oral iron requirement. Retrospective. Positive capsule findings. Rebleeding defined as melena/ hematochezia with drop in hemoglobin, or need for transfusion. Retrospective study of 113 patients undergoing capsule. Of 10 arteriovenous malformation patients, 7 underwent observation. Rebleeding defined as overt bleeding or fall of hemoglobin >2 g/dL compared with baseline in absence of other cause of decline in hemoglobin
*47/94 = 50.0%, if Richter et al36 (including colonic angioectasia) is excluded.
by intraoperative endoscopy (23.5%; 95% CI, 7%-50%), albeit with a wide CI (Table 1). The type of endoscopic therapy was not detailed in some studies, but the 7 studies in Table 2 that reported using contact cautery (without argon plasma coagulation or clips) had similar rebleeding rates to the overall rate (29/78 = 37%), as did the 6 studies that mentioned using clips (30/79 = 38%).
Confounding of Overall Rebleeding Rates by Lesion Mix and Proportion Treated Surgically Overall rebleeding rates have often been used to imply a benefit to endoscopic therapy in obscure bleeding. However, not all lesions in obscure GI bleeding studies are
FIGURE 1. Scatter plot of the rebleed rates from the natural history [no treatment (left)] studies, the rebleed rates after endoscopic therapy at various types of endoscopy access (right), and the rebleed rates after surgical therapy after intraoperative endoscopy localization (middle).
Copyright
r
angioectasias, and case-mix regarding vascular and nonvascular (tumors/ulcers) lesions is highly variable, so these overall rebleed rates can be misleading. Case-mix is important as lesion type appears to predict response; angioectasia lesions respond less favorably to therapy (vs. ulcers or tumors), particularly to medical/endoscopic therapy, in multiple studies.15,18–20,40 The angioectasia patients accounted for 90% of rebleeders after capsule-directed treatment in another study.29 In Gerson et al,20 57% of the 40 who had endoscopic therapy for a vascular lesion rebled in the first year, whereas rebleeding rates for ulcers and tumors were lower after their lesion-specific therapy. In a prospective study (randomized trial of push-first vs. capsulefirst), rebleeding was higher (50% to 63%) for angioectasia specifically, than for all pathologies (30% to 42%).41 Age and type of therapy can also confound overall rates, making them misleading estimates of rebleeding in the average angioectasia patient with bleeding. Surgery was a much more effective treatment for angioectasia (24% vs. 53%; 3-y rebleeding in Richter et al 36) in a few studies (albeit with possible selection biases),19,38 and a lower rate (24% vs. 43%) was seen for the studies we summarized on surgery after intraoperative endoscopy localization (Table 1); the proportion of patients treated surgically can therefore confound overall rebleeding rates, causing them to be poor estimates of the rebleeding rate after endoscopic therapy.19,40 The effect of age on rebleeding is unclear, but is partly explained by the different etiologies found, fewer comorbidities, and different rates of surgery in different age groups. Regardless of the reason, young age seems to predict lower rebleeding rates (0% rebleeding under age 50, vs. 57% otherwise).40 Sun et al42 had an 11% rebleeding rate in a cohort with a relatively young mean age of 48 years. These rates may not be applicable to the elderly patient with angioectasia. The rebleeding rates’ dependency on etiology, age, and rates of surgical therapy (for tumors or angioectasia) may explain why some studies report overall rates of rebleeding
2014 Wolters Kluwer Health, Inc. All rights reserved.
www.jcge.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
827
J Clin Gastroenterol
Romagnuolo et al
after capsule-directed or enteroscopy-directed therapy in obscure bleeding as low as 10%; the lower rebleeding rate studies18,28,32 tend to contain younger patients, a broader mixture of pathologies (more tumors and ulcers and fewer angioectasias), and a higher proportion of patients treated surgically.
DISCUSSION To interpret the postendoscopic therapy rebleeding rates in the literature, and estimate the absolute risk reduction of endoscopic therapy, the natural history of angioectasia lesions needs to be summarized. Most therapeutic studies have been uncontrolled and there are no detailed data on whether rebleeding was over single or multiple occurrences in published series. Much of the rebleeding rate estimates come from studies of diagnostic yield with insufficient extractable data on etiology-specific and treatment-specific outcomes. Summarizing the studies that have this level of detail, and comparing their outcomes to natural history studies, it appears that the absolute risk reduction may be small (6% to 7%) and number need to treat is high (needed to treat 15 to 16); the scatter plot of rebleed rates, with and without therapy (Fig. 1) suggest little if any treatment effect. This may seem counterintuitive; however, recent randomized controlled outcomes trials of small bowel barium imaging versus capsule, and push enteroscopy versus capsule, both nicely illustrated the disconnection between higher diagnostic yield (much higher in capsule) and unchanged outcomes (no difference in rebleeding).41,43 These
Is Endoscopic Therapy Effective for Angioectasia in Obscure Gastrointestinal Bleeding? A Systematic Review of the Literature Joseph Romagnuolo, MD, MSc, FACG, FASGE,* Andrew S. Brock, MD,w and Nathaniel Ranney, MDw
Goal: We aimed to summarize pooled rebleeding rates of angioectasia after therapeutic endoscopy, and compare these to historical control (no intervention) rates. Background: Obscure gastrointestinal bleeding continues to be challenging to diagnose and treat; in America, small bowel angioectasias are the most common cause. Technology advances led to higher diagnostic yield for these lesions; however, therapeutic impact of endoscopy remains unclear. Study: A PubMed search (June 1, 2006 to September 19, 2013) with 2 independent reviews sought articles reporting rebleeding rates of symptomatic angioectasia without therapy (natural history) and after endoscopic treatment. This study list was added to studies in the 2007 American Gastroenterological Association systematic review. Data on number of patients who underwent endoscopic therapy, type of therapy used, number of patients who experienced rebleeding, and follow-up time were extracted. Rebleeding data were pooled and weighted averages were reported with 95% confidence intervals (CI). Results: Twenty-four articles (n = 490 patients) with data on endoscopic therapy for angioectasia and 6 natural history cohorts (n = 130) receiving no therapy for angioectasia were eligible. Of the endoscopic therapy patients, 121 at push enteroscopy and 427 at balloon-assisted enteroscopy; 209/490 (42.7%; 95% CI, 38%-47%) rebled. Of the control (no therapy) patients, 64/130 (49.2%; 95% CI, 40%-58%) rebled. Number needed to treat is estimated at 15 to 16. Conclusions: Rebleeding rate after endoscopic therapy for symptomatic small bowel angioectasia may be comparable to that expected without therapy. Endoscopic therapy may be ineffective; if effective, the needed to treat is estimated to be high. Controlled studies, with intervention-stratified and etiology-stratified outcomes are needed. Key Words: obscure GI bleeding, angioectasia/angiodysplasia, rebleeding, endoscopic therapy, outcomes, natural history, small bowel/intestine
(J Clin Gastroenterol 2015;49:823–830)
Received for publication May 15, 2014; accepted October 9, 2014. From the *Department of Public Health Sciences, Medical University of South Carolina; and wDepartment of Medicine, Division of Gastroenterology and Hepatology, Medical University of South Carolina, Charleston, SC. Presented (by J.R.) in part at ACG 2009 and ACG 2011 at topic forums: “Is endoscopic therapy effective for angioectasia?” and “Endoscopic therapy for angioectasia: are we treating ourselves or the patient?,” respectively. J.R.: concept and design; J.R., A.S.B., N.R.: data collection and analysis; manuscript drafting; critical revision of manuscript. J.R. was funded by an ASGE Senior Investigator Mentoring Award. The remaining authors declare that they have nothing to disclose. Reprints: Joseph Romagnuolo, MD, MSc, FACG, FASGE, Palmetto Health, Columbia Gastroenterology Associates, 2739 Laurel St, Ste 1A, Columbia, SC 29204 (e-mail: [email protected]). Copyright r 2014 Wolters Kluwer Health, Inc. All rights reserved.
J Clin Gastroenterol
A
n angioectasia is a collection of ectatic blood vessels made of a thin wall with or without an endothelial lining, and with or without coexisting arteriovenous communications.1 It is also referred to as arteriovenous malformation, vascular lesion, venous ectasia, angioma, or angiodysplasia. It is most often incidentally found at endoscopy, and 1 g/dL. Retrospective. 133 of 261 had vascular lesions, 129 treated endoscopically (APC), with 31 lost to follow-up. Most (121 of 129) had angioectasia (5 Dieulafoy, 3 Blue Rubber Bleb Nevus syndrome). Rebleeding defined as reappearance of overt bleeding, refractory or iron deficiency anemia, or continued need for transfusion. Retrospective. 5 of 75 had angioectasia, treated endoscopically (APC, epinephrine injection, and/ or clip). Rebleeding defined as hematochezia, melena, or need for blood transfusion. Retrospective. 79 of 147 had angioectasia, treated endoscopically (details not given), with 16 lost to follow-up (8 had lesions “proximal to the ligament of Treitz” which had 38% rebleeding). Rebleeding defined as overt bleeding, hospitalization, need for further intervention, or need for blood transfusion or iron therapy. Retrospective. 10 of 113 had angioectasia, 3 treated endoscopically (APC). Rebleeding defined as overt bleeding or otherwise unexplained hemoglobin drop >2 g/dL compared with baseline. Retrospective. 26 of 39 had angioectasia, treated endoscopically (APC or heater probe). Rebleeding not well defined.
*202/478 = 42.3% if the 4 studies with n < 5 excluded. APC indicates argon plasma coagulation (noncontact coagulation).
smallest (n = 5)39study is excluded] (Table 3, Fig. 1). The control groups did not appear to be comprised of low-risk individuals. In fact, half of the patients in 1 trial38 had lesions prominent enough to be seen on angiography. In addition, most patients enrolled in the 6 studies had a history of recurrent/chronic bleeding rather than a single episode; single episodes were also often hemodynamically significant. The most powerful natural history data are likely that of the randomized trial of hormonal therapy versus placebo, wherein the rebleeding rate without therapy was only 46% at 2 years, despite high-risk presentations.38 The weighted summary rebleeding rate, pooling data from the 6 studies listed, was 64 bleeds in 130 patients or 49.2% (95% CI, 40%-58%), over 1.5 to 2 years.
Rebleeding After Endoscopic Treatment of Small Bowel Angioectasia The rebleeding rates after endoscopic treatment are more difficult to tease out. Most trials have centered on the diagnostic test rather than the treatment, and so, stratified
rebleeding rates (treated vs. untreated, endoscopic vs. other treatment) are difficult to extract. More than half of studies involving enteroscopy do not report rebleeding rates. Many do not report angioectasia-specific outcomes, or describe the therapy for angioectasias. In the papers that do report outcomes after endoscopic therapy for angioectasia patients, the corresponding rebleeding rates were variable from study to study, and from procedure to procedure (Table 2, Fig. 1). With push enteroscopy, the range in rebleeding rates was 0% to 66%, with a weighted mean of 34.9% (6 studies, n = 63). With deep (balloon-assisted) enteroscopy, the range in rebleeding rates was 24% to 67%, with a weighted mean of 43.8% (18 studies, n = 427). The pooled data from these 24 studies included 490 patients, of which 209 patients rebled after endoscopic intervention, for an overall weighted rebleed rate of 42.7% (95% CI, 38%-47%), over 1.5 to 2 years. This estimate has CIs that overlap markedly with those of the natural history pooled rate; it is also higher than that from the surgical resection studies after lesion identification
826 | www.jcge.com Copyright r 2014 Wolters Kluwer Health, Inc. All rights reserved. Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
J Clin Gastroenterol
Volume 49, Number 10, November/December 2015
Effectiveness of Endoscopy for Angioectasia
TABLE 3. Natural History Studies of Rebleeding of Angioectasia
n
Rebleeding Rate n (%)
Average/median Follow-up (y)
36
17 (47)
3
Lewis et al37
34
19 (56)
1
Junquera et al38
35
16 (46)
2
Saurin et al30
13
6 (46)
1
Park et al39
5
4 (80)
2.6
Koh et al34
7
2 (28)
2 (mean)
64 (49.2)/95% confidence interval, 40%-58%*
1.9
References Richter et
al36
Totals
130
Comments Retrospective. Medically treated. Colonic angioectasias included.* Rebleeding defined as needing transfusion, hospitalization, or surgery. Retrospective. Patients who refused hormonal therapy or prehormonal therapy era. Rebleeding defined as needing transfusion. Placebo arm of randomized trial of hormonal therapy for angioectasia diagnosed by endoscopy and angiography. Rebleeding defined as acute bleed vs. anemia refractory to iron therapy with positive fecal occult blood Retrospective analysis of previous prospective study comparing push endoscopy and capsule endoscopy. Rebleeding defined as presence of persistent anemia, transfusion requirement, or oral iron requirement. Retrospective. Positive capsule findings. Rebleeding defined as melena/ hematochezia with drop in hemoglobin, or need for transfusion. Retrospective study of 113 patients undergoing capsule. Of 10 arteriovenous malformation patients, 7 underwent observation. Rebleeding defined as overt bleeding or fall of hemoglobin >2 g/dL compared with baseline in absence of other cause of decline in hemoglobin
*47/94 = 50.0%, if Richter et al36 (including colonic angioectasia) is excluded.
by intraoperative endoscopy (23.5%; 95% CI, 7%-50%), albeit with a wide CI (Table 1). The type of endoscopic therapy was not detailed in some studies, but the 7 studies in Table 2 that reported using contact cautery (without argon plasma coagulation or clips) had similar rebleeding rates to the overall rate (29/78 = 37%), as did the 6 studies that mentioned using clips (30/79 = 38%).
Confounding of Overall Rebleeding Rates by Lesion Mix and Proportion Treated Surgically Overall rebleeding rates have often been used to imply a benefit to endoscopic therapy in obscure bleeding. However, not all lesions in obscure GI bleeding studies are
FIGURE 1. Scatter plot of the rebleed rates from the natural history [no treatment (left)] studies, the rebleed rates after endoscopic therapy at various types of endoscopy access (right), and the rebleed rates after surgical therapy after intraoperative endoscopy localization (middle).
Copyright
r
angioectasias, and case-mix regarding vascular and nonvascular (tumors/ulcers) lesions is highly variable, so these overall rebleed rates can be misleading. Case-mix is important as lesion type appears to predict response; angioectasia lesions respond less favorably to therapy (vs. ulcers or tumors), particularly to medical/endoscopic therapy, in multiple studies.15,18–20,40 The angioectasia patients accounted for 90% of rebleeders after capsule-directed treatment in another study.29 In Gerson et al,20 57% of the 40 who had endoscopic therapy for a vascular lesion rebled in the first year, whereas rebleeding rates for ulcers and tumors were lower after their lesion-specific therapy. In a prospective study (randomized trial of push-first vs. capsulefirst), rebleeding was higher (50% to 63%) for angioectasia specifically, than for all pathologies (30% to 42%).41 Age and type of therapy can also confound overall rates, making them misleading estimates of rebleeding in the average angioectasia patient with bleeding. Surgery was a much more effective treatment for angioectasia (24% vs. 53%; 3-y rebleeding in Richter et al 36) in a few studies (albeit with possible selection biases),19,38 and a lower rate (24% vs. 43%) was seen for the studies we summarized on surgery after intraoperative endoscopy localization (Table 1); the proportion of patients treated surgically can therefore confound overall rebleeding rates, causing them to be poor estimates of the rebleeding rate after endoscopic therapy.19,40 The effect of age on rebleeding is unclear, but is partly explained by the different etiologies found, fewer comorbidities, and different rates of surgery in different age groups. Regardless of the reason, young age seems to predict lower rebleeding rates (0% rebleeding under age 50, vs. 57% otherwise).40 Sun et al42 had an 11% rebleeding rate in a cohort with a relatively young mean age of 48 years. These rates may not be applicable to the elderly patient with angioectasia. The rebleeding rates’ dependency on etiology, age, and rates of surgical therapy (for tumors or angioectasia) may explain why some studies report overall rates of rebleeding
2014 Wolters Kluwer Health, Inc. All rights reserved.
www.jcge.com |
Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.
827
J Clin Gastroenterol
Romagnuolo et al
after capsule-directed or enteroscopy-directed therapy in obscure bleeding as low as 10%; the lower rebleeding rate studies18,28,32 tend to contain younger patients, a broader mixture of pathologies (more tumors and ulcers and fewer angioectasias), and a higher proportion of patients treated surgically.
DISCUSSION To interpret the postendoscopic therapy rebleeding rates in the literature, and estimate the absolute risk reduction of endoscopic therapy, the natural history of angioectasia lesions needs to be summarized. Most therapeutic studies have been uncontrolled and there are no detailed data on whether rebleeding was over single or multiple occurrences in published series. Much of the rebleeding rate estimates come from studies of diagnostic yield with insufficient extractable data on etiology-specific and treatment-specific outcomes. Summarizing the studies that have this level of detail, and comparing their outcomes to natural history studies, it appears that the absolute risk reduction may be small (6% to 7%) and number need to treat is high (needed to treat 15 to 16); the scatter plot of rebleed rates, with and without therapy (Fig. 1) suggest little if any treatment effect. This may seem counterintuitive; however, recent randomized controlled outcomes trials of small bowel barium imaging versus capsule, and push enteroscopy versus capsule, both nicely illustrated the disconnection between higher diagnostic yield (much higher in capsule) and unchanged outcomes (no difference in rebleeding).41,43 These
