Do Clinical Characteristics Predict the Presence of Small Bowel Angioectasias on Capsule Endoscopy?
- First Online:
- Cite this article as:
- DeBenedet, A.T., Saini, S.D., Takami, M. et al. Dig Dis Sci (2011) 56: 1776. doi:10.1007/s10620-010-1506-9
- 146 Views
Background and Aims
Angioectasias (AVMs) are the most common vascular anomaly of the gastrointestinal (GI) tract, and these lesions are often associated with obscure gastrointestinal bleeding (OGIB). It is unknown if the presence of upper and/or lower gastrointestinal AVMs are predictive of small bowel AVMs. The aims of this study are to define the small bowel segmental distribution of AVMs and to identify the factors predicting the presence of small bowel AVMs among a cohort of patients with a known history of AVMs in the upper and/or lower GI tracts who are undergoing capsule endoscopy (CE) for OGIB.
We performed a retrospective cohort analysis of 1,125 patients undergoing CE at our institution between 11/1/2001 and 8/31/2007. Inclusion criteria were: (1) complete esophagoduodenoscopy (EGD), CE, and colonoscopy, (2) OGIB indication for CE, and (3) history of ≥1 AVM on EGD and/or colonoscopy that was previously treated in the past or deemed not to be a clinically significant source of bleeding. Exclusion criteria were: (1) history of radiation therapy to the GI tract, and (2) presence of a congenital or systemic disease associated with GI AVMs. Data were extracted on: (1) age; (2) gender; (3) presence of diabetes, (4) presence of hypertension, (5) presence of aortic stenosis, (6) history of non-steroidal anti-inflammatory therapy, (7) history of anticoagulant therapy, (8) hemoglobin, platelet, and INR values prior to CE; (9) baseline serum creatinine; and (10) presence and GI tract segmental location of AVMs. Multivariate logistic regression was used to identify independent predictors of small bowel AVMs.
1,125 patients underwent EGD, CE, and colonoscopy. One hundred and fourteen patients had a history of ≥1 AVM on EGD and/or colonoscopy and met inclusion and exclusion criteria. The mean age was 69 years, and 63% of patients were women. 37% of patients were found to have ≥1 jejunal AVM and 15% were found to have ≥1 ileal AVM. In multivariate analysis, age ≥ 65 (OR 2.62, P = 0.05) and the presence of AVMs on EGD (OR 4.61, P = 0.02) were predictive of jejunal AVMs. AVMs on colonoscopy alone were not predictive of jejunal or ileal AVMs. No factors were found to predict the presence of ileal AVMs.
Patients with AVMs on EGD have an increased risk of jejunal AVMs on CE, particularly if they are elderly. Future studies should validate these findings in a prospective cohort.
KeywordsAngioectasiasObscure gastrointestinal bleedingCapsule endoscopyDouble-balloon enteroscopyOGIBAVM
American gastroenterological association
von Willebrand’s factor
Angioectasias (AVMs), also known as angiodysplasias, arteriovenous malformations, and vascular ectasias, are the most common vascular anomalies of the gastrointestinal (GI) tract. These lesions are composed of dilated, ectatic, thin-walled vessels—typically submucosal veins—found in the mucosa and submucosa . The pathophysiology of AVMs, excluding those lesions resulting from congenital diseases, systemic diseases, or radiation therapy, is unclear; however, several theories have been proposed, including acute and chronic mucosal ischemia [2, 3], chronic low-grade venous obstruction , and genetic factors.
AVMs are a common source of obscure gastrointestinal bleeding (OGIB). Yet, because many patients are asymptomatic, the exact prevalence of AVMs is unknown. It is estimated that only 10% of all patients with AVMs have clinically evident GI bleeding during their lifetime , but these lesions account for up to 80% of all cases of obscure bleeding in patients with ongoing transfusion requirements . Some important factors have been shown to be associated with a greater likelihood of gastrointestinal bleeding from AVM(s). For instance, in elderly populations, AVMs comprise up to 40% of all causes of small bowel bleeding [4, 5]. Furthermore, AVMs are an important cause of GI bleeding in patients with end-stage renal disease, possibly due to uremia-induced platelet dysfunction [6, 7]. An association between bleeding from gastrointestinal AVMs and von Willebrand’s disease has also been described [8–10]. Similarly, von Willebrand’s factor consumption by a stenotic aortic valve may also explain the association between the presence of aortic stenosis and increased GI bleeding from AVMs [11–15].
Though AVMs are an important cause of obscure GI bleeding, the optimal approach to diagnosing and treating these lesions remains unclear. Much of this uncertainty stems from our limited knowledge of the segmental distribution of these lesions and a lack of clinical predictors for the presence of AVMs in the GI tract. Data suggest that AVMs may cluster within intestinal segments , and it is estimated that 40–60% of patients have more than one lesion [17, 18]. Further, the reported prevalence of concurrent upper and lower GI tract lesions varies from “rare” to as high as 33% [16, 19–22]. As a result of this uncertainty, in patients with OGIB and known, but either previously treated or clinically insignificant upper and/or lower GI tract AVMs, the most appropriate diagnostic “next step” is unclear.
The purpose of this study was to define the segmental distribution of small bowel AVMs in patients with known AVMs on EGD and/or colonoscopy. Secondly, we attempted to determine whether certain clinical characteristics are predictive of the presence of small bowel AVMs, specifically characteristics that are known to increase the likelihood of AVM bleeding.
Materials and Methods
Ascertainment of Cohort Sample
Data Collection and Management
Data were extracted on: (1) age; (2) gender; (3) presence of diabetes, (4) presence of hypertension, (5) presence of aortic stenosis, (6) history of non-steroidal anti-inflammatory therapy, (7) history of anticoagulant therapy, (8) hemoglobin, platelet, and INR values prior to CE; (9) baseline serum creatinine; (10) presence of AVMs on EGD, colonoscopy, and/or CE; and, (11) location of AVMs (gastric, duodenal, jejunal, ileal, and/or colonic). To accomplish a standardized, accurate, and reproducible electronic medical record review at all stages of our study, we utilized a computer program developed at our institution known as “EMERSE ” which electronically and automatically searched the entire electronic medical record for our data of interest. EMERSE, an acronym for “The Electronic Medical Record Search Engine” employs a similar search algorithm to the well known internet search engine, Google . The user submits a search term or phrase into the search field and EMERSE parses the electronic medical records for the specified patient to find the term(s) of interest. EMERSE then creates a listing of documents where the term(s) appear and specifically generates an output summary page that contains the sentence where the term(s) appear in each document. EMERSE further enables the user to search and cross-reference all of the documents that contain the term(s) for complete data extraction. Microsoft Excel spreadsheets were used for data management.
Aortic stenosis was defined as present if the stenosis was graded as “mild” or greater according to recent guidelines . Baseline serum creatinine was defined as the mean of the recorded serum creatinine values within 3 months of the CE. Exceptions to this definition included: (1) instances of acute renal failure, where the serum creatinine values during the acute renal failure period were not included; (2) instances of only a single serum creatinine measurement, in which case the 3-month time interval was expanded to 6 months; (3) patients on dialysis (peritoneal or hemodialysis), who were categorized as “dialysis dependent.” History of NSAID therapy (including aspirin) was defined as taking at least 1 pill per day for greater than 1 week within 1 month prior to CE. History of anti-coagulant or anti-platelet therapy (i.e. clopidogrel, warfarin, etc.) was defined as taking at least 1 pill within 1 week of CE. CE images were reviewed by one of two CE experts at our institution. AVM presence and location was determined by review of the CE report. For the purpose this study, jejunal location was defined as the area visualized in the first half of the capsule study, between the first duodenal image and the first cecal image. Lesions seen in the duodenum proximal to the ampulla, were excluded.
The primary outcomes of interest were the presence of jejunal AVMs or the presence of ileal AVMs. The predictor variables of interest were presence of AVMs on EGD and/or colonoscopy. Covariates included age, gender, chronic kidney disease (creatinine or hemodialysis dependency), and aortic stenosis. The remainder of the extracted data was utilized as descriptive characteristics for our cohort. Logistic regression was used to identify individual (bivariate) predictors of the presence of jejunal or ileal lesions. Covariates found to be potentially important in bivariate analysis (P-value < 0.10) were included in a multivariate logistic regression analysis that included the presence of AVMs on EGD or colonoscopy. Adjusted odds ratios and 95% confidence intervals were presented. An a priori determined P-value of ≤0.05 was considered statistically significant. All analyses were performed using the Stata 10.0 statistical package (Statacorp, College Station, Texas).
Number of patients
Mean age ± SD
70 ± 11
History of NSAID therapy
History of anti-coagulant therapy
Mean hemoglobin ± SD
10.7 ± 1.6
Mean platelets ± SD
262 ± 154
Mean INR ± SD
1.3 ± 0.6
Baseline serum creatinine
1.14 ± 0.44
AVM(s) on EGD
AVM(s) on CE
AVM(s) on colonoscopy
Segmental location of AVMs
Proportion of study cohort with AVM
Bivariate analysis: predictors of jejunal AVMs
Age ≥ 65
Baseline serum creatinine
≥1 AVM(s) on EGD
≥1 AVM(s) on colonoscopy
Multivariate analysis: predictors of jejunal AVMs
Age ≥ 65
≥1 AVM(s) on EGD
≥1 AVM(s) on Colonoscopy
Though AVMs are an important cause of obscure GI bleeding, the optimal approach to diagnosing and treating these lesions remains unclear. Much of this uncertainty stems from our limited knowledge of the segmental distribution of these lesions and a lack of clinical predictors for the presence of AVMs in the GI tract. As a result, the most appropriate “next step” in the management of patients with histories of AVMs on EGD or colonoscopy and recurrent bleeding is not well defined. Our study suggests that the most common location for small bowel AVMs is the jejunum. Furthermore, age ≥65 and the presence of AVMs on EGD were found to be predictive of jejunal AVMs but not ileal AVMs. There were no clinical characteristics predictive of ileal AVMs, and AVMs on colonoscopy alone were not predictive of either jejunal or ileal AVMs.
Some important limitations of our study should be highlighted. This was a retrospective, hypothesis-generating study with patients from an academic referral population who had undergone complete visualization of the GI tract. Although the presence and location of AVMs was well documented, the clinical significance of each lesion as a definite source of blood loss could not be determined. Further, the clinical characteristics assessed in our statistical models—namely age, gender, baseline serum creatinine, and the presence of aortic stenosis—were chosen based on their reported associations with an increased likelihood of bleeding from AVMs. It is possible that there are other characteristics or factors, currently unknown, which influence the risk of having AVMs that were not included, and thus not controlled for, in our study. Although data were extracted on history of NSAID and anti-coagulant therapies, these therapies were not included as variables in our model. NSAIDS and anticoagulant therapy have been associated with longer hospital stays in GI bleeders  and with increased rebleeding rates in patients with supratherapeutic anticoagulant treatment . However, the precise association between these therapies and the increased presence of AVMs or increased risk of bleeding from AVMs needs further clarification These caveats may limit the generalizability of our findings, although our AVM prevalences are consistent with reported data. Finally, ileal AVMs occurred infrequently in our cohort (17 patients), limiting our ability to draw robust statistical conclusions about predictors of these lesions.
Currently, the American Gastroenterological Association (AGA) and the American Society for Gastrointestinal Endoscopy (ASGE) recommend CE as the first line diagnostic test for evaluation of the small bowel in patients with obscure gastrointestinal bleeding if no contraindications exist [29, 30]. CE has a higher diagnostic yield for clinically significant lesions than small bowel enteroscopy  and high sensitivity compared to the gold standard of intraoperative enteroscopy , thus allowing for possible earlier diagnosis . In a meta-analysis comparing DBE with VCE in patients with OGIB, the pooled yield for AVMs was identical for both technologies . DBE or other overtube-assisted deep bowel enteroscopy techniques may convey additional advantage in patients with obscure GI bleeding at risk for small bowel AVMs, as this procedure allows for therapeutic intervention and may therefore be a cost-effective initial approach when compared to initial CE . If future data confirm these findings, initial DBE or push enteroscopy (depending on availability of procedure and expertise of the providers) may be a reasonable option in selected patients with obscure GI bleeding suspected of having small bowel AVMs.
In conclusion, this study demonstrated that patients with AVMs on EGD were at increased risk for jejunal AVMs on CE, particularly if they were elderly. Future studies should validate these findings in a prospective cohort. Patients who satisfy these criteria may benefit more from immediate therapy with balloon assisted enteroscopy than from further diagnostic testing. The cost-effectiveness of proceeding directly to balloon assisted enteroscopy in these high-risk patients merits formal evaluation.
We would like to thank Brian Burtraw, Mark McDermott, and Jason Baker (University of Michigan Health System Gastrointestinal Physiology Laboratory Technicians) for their assistance in the management of our CE data.
Conflict of interest
The authors declare no conflicts of interests associated with the creation of this manuscript.