Predictors of active and recurrent bleeding in upper gastrointestinal angiodysplasia: Results from 10-year esophagogastroduodenoscopy of a tertiary center in Taiwan Authors: Meng-Chuan Lu, Tien-Yu Huang, Peng-Jen Chen, Hsuan-Wei Chen, Chih-Wei Yang, Jung-Chun Lin, Wei-Chen Huang, Hsuan-Hwai Lin, Yu-Lueng Shih, Tsai-Yuan Hsieh, Bao-Chung Chen Published: 2024 ABSTRACT Upper gastrointestinal angiodysplasia (UGIA) is a unique mucosal vascular lesion that causes acute or recurrent gastrointestinal bleeding. Despite the increasing incidence of UGIA, the risk factors for bleeding in this condition remain unclear. We investigated the predictors of active and recurrent bleeding among patients with UGIA. We conducted a retrospective cohort study of 104,086 patients who underwent esophagogastroduodenoscopy (EGD) at the Tri-Service General Hospital in Taiwan between October 2011 and September 2021. UGIA was diagnosed in 112 patients, and the data from these patients were comprehensively analyzed. EGD documented UGIA in 112 patients, which corresponded with an overall diagnostic yield of 1.1%. Active bleeding was documented in 28 (25%) patients. Patients in the active bleeding group were older than the patients in the asymptomatic group (71.6 ± 11.7-years vs 61.3 ± 18.5-years). Most of the UGIA lesions were singular and measured 2 to 5 mm. Nearly 1-quarter of the patients (24.1%) underwent endoscopic hemostasis, and rebleeding occurred in 6 (21.4%) patients during the follow-up period. Multivariate analysis identified diabetes mellitus (DM), chronic kidney disease (CKD), and lesions in the duodenum as significant risk factors for active bleeding (P = .019; P = .006; P = .004). Our cohort study provided real-world data on the clinical and endoscopic features of UGIA. DM, CKD, and lesions in the duodenum were independent predictors of active bleeding. 1. Introduction Upper gastrointestinal angiodysplasia (UGIA) is a unique mucosal vascular lesion that causes acute or recurrent gastrointestinal bleeding. Its etiology and incidence remain unclear, but its pathogenesis is believed to be similar to colonic angiodysplasia.[1] Several studies have demonstrated that angiodysplasia begins with chronic, intermittent, low-grade obstruction of the submucosal veins, which results in dilation and tortuosity of the submucosal veins, venules, capillaries, and arteries of the mucosal vascular unit. The precapillary sphincters become dysfunctional, and its fragile vascular network becomes prone to rupture.[2,3] UGIA accounts for approximately 1.2% to 8% of patients with UGI bleeding, and the incidence of UGIA is increasing because of improvements in endoscopic screening and imaging techniques.[4,5] UGIA demonstrates 3 patterns on endoscopy. Pattern I is the most common presentation and is characterized by flat or slightly elevated, bright-red lesions with frond-like margins, whereas patterns II and III are characterized by telangiectatic and submucosal nodular lesions, respectively.[6] The epidemiology of UGI bleeding has changed over the last decade. The prevalence of gastritis and peptic ulcer disease in patients with UGI bleeding decreased by 55% and 30%, respectively, but the prevalence of angiodysplasia increased by 30%.[7] Some groups have reported that angiodysplasia of the GI tract is associated with age, chronic renal failure (CRF), von Willebrand disease (vWD), aortic stenosis, and left ventricular assist devices.[8–12] Despite the increasing incidence of UGIA in clinical practice, there are limited data on the risk factors for bleeding in UGIA.[13] This study examined the predictors of active and recurrent bleeding among patients with UGIA. 2. Materials and methods 2.1. Study design and population We retrospectively analyzed the data of 104,086 consecutive patients who underwent esophagogastroduodenoscopy (EGD) at the Tri-Service General Hospital in Taiwan between October 2011 and September 2021. UGIA was documented in 112 patients. Patients with hereditary hemorrhagic telangiectasia, gastric antral vascular ectasia, angioma, arteriovenous malformation, radiation-induced vascular ectasia, and Dieulafoy’s lesion were excluded from this study.[14] Active bleeding was defined as spurting or oozing of blood from a UGIA lesion during EGD, whereas recurrent bleeding was defined as hemorrhage following an episode of active bleeding. Chronic kidney disease (CKD) was defined as an estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73 m2. The study design was approved by the Institutional Review Board of Tri-Service General Hospital (TSGHIRB 1-106-05-154). 2.2. Data collection Demographic and clinical data were collected from medical charts and reviewed. Clinical data included comorbidities, endoscopic photographs, and clinical outcomes. Follow-up data were collected until December 2022. 2.3. Statistical analysis Statistical analyses were performed on SPSS software version 22 (IBM, Chicago, Illinois, USA). Normally distributed data were presented as mean ± standard deviation. The analysis of ANOVA, chi-square test, and Fisher exact test were used for continuous and categorical variables, as necessary. Logistic regression analysis was utilized to compare the incidence of bleeding between the active bleeding and asymptomatic groups. Statistical significance was set at P value < .05. 3. Results 3.1. Demographic and clinical features of the study population The demographic and clinical characteristics of the study population are shown in Table 1. Patients were categorized into the active bleeding and asymptomatic groups according to the flow chart in Figure 1. In the active bleeding group, the mean age was 71.6 ± 11.7-years, and the majority of patients were male (64.3%). In contrast, the asymptomatic group had a younger mean age of 61.3 ± 18.6-years, and most of the patients were female (56%). The most common comorbidities in both groups were hypertension, diabetes mellitus (DM), CKD, and coronary artery disease; 60.7%, 50%, 50%, and 25% of the patients in the active bleeding group had these conditions, respectively, whereas 35.7%, 14.3%, 10.7%, and 10.7% of the patients in the asymptomatic group had these conditions, respectively. The proportion of patients with antiplatelet or anticoagulant drug, cigarette smoking, and alcohol use disorder was higher in the active bleeding group than in the asymptomatic group (35.7% vs 17.9%, Odd ratio = 2.556, 95% CI: 1.011 to 6.632, P value = 0.049; 35.7% vs 21.4%, Odd ratio = 2.037, 95% CI: 0.802 to 5.175, P value = 0.131; 10.7% vs 4.8%, Odd ratio = 2.4, 95% CI: 0.503 to 11.454, P value = 0.364, respectively). Table 1Clinical features of the 112 patients with UGIA.*Active bleeding group (%)Asymptomatic group (%)Odd ratioP valueConfidence interval (CI)Age (yr)71.6 ± 11.761.3 ± 18.6.007 Age ≧ 6521 (75)37 (44)3.811.0051.462 to 9.93Sex(male/female)18/10 (64.3/35.7)37/47 (44/56)2.286.0640.944 to 5.54Comorbid illness Hypertension17 (60.7)30 (35.7)2.782.021.154 to 6.707 Coronary artery disease7 (25)9 (10.7)2.778.1140.925 to 8.344 Valvular heart disease4 (14.3)5 (6)2.633.2240.655 to 10.592 Atrial fibrillation5 (17.9)6 (7.1)2.826.1390.79 to 10.11 Diabetes mellitus14 (50)12 (14.3)602.297 to 15.676 Cirrhosis of liver5 (17.9)5 (6)3.435.1170.914 to 12.906 Chronic kidney disease (eGFR† < 60mL/min/1.73m2)14 (50)9 (10.7)8.33303.025 to 22.956 Antiplatelet or anticoagulant drug use10 (35.7)15 (17.9)2.556.0491.011 to 6.632 Cigarette smoking10 (35.7)18 (21.4)2.037.1310.802 to 5.175 Alcohol use disorder3 (10.7)4 (4.8)2.4.3640.503 to 11.454*UGIA = upper gastrointestinal angiodysplasia.†eGFR = estimated glomerular filtration rate. Clinical features of the 112 patients with UGIA.* UGIA = upper gastrointestinal angiodysplasia. eGFR = estimated glomerular filtration rate. Allocation of patients enrolled in this study. Abbreviations: EGD = esophagogastroduodenoscopy; UGIA = upper gastrointestinal angiodysplasia. The percentage of UGIA patients in different age groups in this study was disclosed in Figure 2. The percentage of UGIA patients in different age groups. Abbreviation: UGIA = upper gastrointestinal angiodysplasia. Univariate analysis demonstrated significant differences in age > 65-years and the prevalence of hypertension, DM, CKD, and antiplatelet or anticoagulant drug use between both groups (Table 1). For those variables with P value < .05 in the study, further multivariable analysis with logistic regression analysis was also conducted. Multivariate analysis revealed significant differences between the prevalence of CKD and DM between both groups (Table 4). 3.2. Endoscopic findings in patients with UGIA EGD diagnosed UGIA in 112 patients. The diagnostic yield of EGD in our study was 1.1%. Digestive tract angiodysplasias are classified according to the number, size, and location of lesions identified during endoscopy.[15] The endoscopic pictures and findings in our patients are shown in Figure 3 and Table 2. Most of the lesions in the active bleeding and asymptomatic groups were singular (78.6% vs 75%) and measured 2 to 5 mm (64.3% vs 65.5%). In the active bleeding group, angiodysplastic lesions were noted in the corpus/fundus, antrum, and duodenum in 10 (35.7%), 4 (14.3%), and 14 (50%) cases, respectively. In the asymptomatic group, angiodysplastic lesions were documented in the esophagus, cardia, corpus/fundus, antrum, and duodenum in 5 (6%), 5 (6%), 50 (59.5%), 9 (10.7%), and 10 (11.9%) cases, respectively. One patient in the asymptomatic group had concurrent UGIA of the esophagus and stomach (1.2%), while 4 patients had concurrent UGIA of the stomach and duodenum (4.8%). Among the 112 patients with UGIA, 27 (24.1%) patients underwent endoscopic hemostasis with argon plasma coagulation (APC) (11.6%), hemostatic grasper application (7.1%), diluted epinephrine injection (1.8%), hemoclipping (0.9%), APC with hemoclipping (0.9%), and APC with hemostatic grasper (0.9%). No serious adverse events, such as perforation or cardiopulmonary distress, were noted during treatment (Table 3). Table 2Endoscopic findings in the 112 patients with UGIA.*EGD† findingsActive bleeding group (%)Asymptomatic group (%)Odd ratioP valueConfidence interval (CI)Number Unique (n = 1)22 (78.6)63 (75)1.222.7020.437 to 3.42 Multiple (n ≧ 2)6 (21.4)21 (25)0.818.7020.292 to 2.289Size Minute (<2 mm)9 (32.1)24 (28.6)1.184.720.47 to 2.982 Intermediate (2 to 5 mm)18 (64.3)55 (65.5)0.949.9090.388 to 2.321 Large (>5 mm)1 (3.6)5 (6)0.58510.065 to 5.234Location Esophagus0 (0)5 (6).329 Cardia0 (0)5 (6).329 Corpus/fundus10 (35.7)50 (59.5)0.378.0290.12 to 0.716 Antrum4 (14.3)9 (10.7)1.389.7340.392 to 4.918 Duodenum14 (50)10 (11.9)7.4<.0012.743 to 19.963 E‡ and S§0 (0)1 (1.2)1 S and D‖0 (0)4 (4.8).57*UGIA = upper gastrointestinal angiodysplasia.†EGD = esophagoduodenoscopy.‡E = esophagus.§S = stomach.‖D = duodenum. Endoscopic findings in the 112 patients with UGIA.* UGIA = upper gastrointestinal angiodysplasia. EGD = esophagoduodenoscopy. E = esophagus. S = stomach. D = duodenum. Table 3Outcomes of the 112 patients with UGIA.*Overall (%)Active bleeding group (%)Asymptomatic group (%)Diagnostic yield112/104,086 (0.11)Endoscopic hemostasis Nil85 (75.9)2 (7.1)83 (98.8) APC†13 (11.6)12 (42.8)1 (1.2) Hemostatic grasper8 (7.1)8 (28.6)0 Hemoclipping1 (0.9)1 (3.6)0 Diluted epinephrine injection2 (1.8)2 (7.1)0 APC with hemoclipping2 (1.8)2 (7.1)0 APC with hemostatic grasper1 (0.9)1 (3.6)0 Rebleeding rate6/112 (5.4)6/28 (21.4)0/84 (0) Adverse event0/112 (0)0/28 (0)0/84 (0)*UGIA = upper gastrointestinal angiodysplasia.†APC = argon plasma coagulation. Outcomes of the 112 patients with UGIA.* UGIA = upper gastrointestinal angiodysplasia. APC = argon plasma coagulation. Endoscopic photographs of UGIA. (A) UGIA in esophagus; (B) UGIA in cardia; (C) UGIA in corpus/fundus; (D) UGIA in antrum; (E) UGIA in duodenal bulb; (F) UGIA in descending part of duodenum. Abbreviation: UGIA = upper gastrointestinal angiodysplasia. Our univariate analysis demonstrated no significant association between the number and size of UGIA lesions and active bleeding (Table 2). There was a significant difference in the location of lesions in the gastric corpus/fundus and duodenum between the active bleeding and asymptomatic groups (P value = .029; P value = 0). Multivariate analysis also revealed a significant difference in the location of lesions in the duodenum between both groups (P = .004) (Table 4). Table 4Multivariate analysis of active bleeding in UGIA.*Variable95% CI†P valueAge ≧ 65-yr0.76 to 8.87.128Hypertension0.11 to 1.77.245Diabetes mellitus1.31 to 19.73.019Chronic kidney disease1.62 to 19.17.006Antiplatelet or anticoagulant drug use0.48 to 6.14.412Lesion in the corpus/fundus0.29 to 5.45.763Lesion in the duodenum2.12 to 54.43.004*UGIA = upper gastrointestinal angiodysplasia.†CI = confidence interval. Multivariate analysis of active bleeding in UGIA.* UGIA = upper gastrointestinal angiodysplasia. CI = confidence interval. 3.3. Rebleeding post-EGD Complete data were collected for all 112 patients, and no patient was lost to follow-up. Recurrent bleeding was noted in 5.4% of patients after an average period of 18 ± 15.2 months from the EGD. The rebleeding rate in the active bleeding group was 21.4%. No bleeding episodes were noted in the asymptomatic group throughout our study duration (Table 3). No statistical significance was found between rebleeding and the clinical or endoscopic factors of age, sex, comorbidities, antiplatelet or anticoagulant on therapy, smoking, alcohol use disorder, the size, number and location of UGIA lesions, or endoscopic hemostasis (Table 5). Table 5Factors influencing rebleeding in the 28 patients in the active bleeding group.Rebleeding (%)Non-rebleeding (%)Odds ratioP valueAge (yr)69 ± 11.772.4 ± 12.544 Age ≧ 654 (66.7)17 (77.3)0.588.622Sex (male/female)4/2 (66.7/33.3)14/8 (63.6/36.4)1.1431Comorbidity Hypertension3 (50)14 (63.6)0.571.653 Coronary artery disease1 (16.7)6 (27.3)0.5331 Valvular heart disease1 (16.7)3 (13.6)1.2671 Atrial fibrillation0 (0)5 (22.7)0.739.553 Diabetes mellitus3 (50)11 (50)11 Cirrhosis of liver1 (16.7)4 (18.2)0.91 Chronic kidney disease5 (83.3)9 (40.9)7.222.165 Antiplatelet or anticoagulant drug use2 (33.3)8 (36.4)0.8751 Cigarette smoking2 (33.3)8 (36.4)0.8751 Alcohol use disorder0 (0)3 (13.6)0.761Number Single (n = 1)3 (50)19 (86.4)0.158.091 Multiple (n ≧ 2)3 (50)3 (13.6)6.333.091Size Minute (<2 mm)3 (50)6 (27.3)2.667.352 Intermediate (2–5 mm)3 (50)15 (68.2)0.467.634 Large (>5 mm)0 (0)1 (4.5)0.7781Location Corpus/fundus2 (33.3)8 (36.4)0.8751 Antrum2 (33.3)2 (9.1)5.191 Duodenum2 (33.3)12 (54.5)0.417.648 Endoscopic hemostasis6 (100)20 (90.9)1.31 Factors influencing rebleeding in the 28 patients in the active bleeding group. 4. Discussion This retrospective cohort study examined 112 patients who were diagnosed with UGIA following EGD of 104,086 consecutive patients in Taiwan. The diagnostic yield of EGD for UGIA in this study was 1.1%, which was lower than the previously reported diagnostic yield of colonoscopy for colonic angiodysplasia (8.3%).[16] The percentage of UGIA patients in different age grouping in this study is shown in Figure 2. Most patients were above 60-years of age, and both men and women were almost equally affected (male to female ratio, 55:57). The active bleeding group was older (71.6 ± 11.7-years vs 61.3 ± 18.6-years) and had more men (64.3% vs 44%) than the asymptomatic group. These were compatible with the findings of previous research.[17,18] Several conditions are associated with angiodysplasia of the GI tract, including old age, CRF, vWD, aortic stenosis, and end-stage heart failure requiring left ventricular assist devices. Several studies have also identified hypertension, arrhythmia, ischemic heart disease, valvular heart disease, liver cirrhosis, DM, and anticoagulant usage as risk factors for GIA.[19–23] Our study revealed that CKD and DM significantly predisposed patients with UGIA to active bleeding (P value = .019; P value = .006). Chalasani et al reported that the prevalence of angiodysplasia was related to the duration and severity of CRF.[24] The pathogenesis of active bleeding in UGIA among patients with CKD may be due to mucosal ischemia provoked by renal failure and linked to other diseases, such as diabetic microangiopathy and atherosclerotic peripheral vascular disease. Hypercalcemia-induced intestinal vascular calcification and impaired microcirculation has also been proposed as the cause of bleeding in UGIA among patients with CKD, whereas platelet dysfunction with abnormal vW factor activity has been considered in patients with uremia.[25] Grooteman et al reported that DM was associated with bleeding in GIA (odds ratio [OR]: 2.6; 95% confidence interval [CI]: 1.03–6.7); however, other studies did not demonstrate a correlation between DM and bleeding in GIA.[13,23,26] Our analysis suggested that DM increased the risk for bleeding in UGIA. It remains unclear whether glycemic control plays a role in the incidence of bleeding in GIA; but hyperglycemia may be an important factor, because it produces reactive oxygen species, which mediate oxidative stress.[27] Further investigation is warranted to establish whether there is a significant relationship between glycemic levels and bleeding in UGIA. Several studies have shown that multiple GIA lesions and GIA lesions smaller than 5 mm or larger than 1 cm were more prone to bleeding.[13,23,28] In our study, the number and size of UGIA lesions were not associated with an increased risk for active bleeding, but UGIA lesions located in the duodenum were more likely to bleed. Some researchers have proposed that muscular contraction is strongest in the gastric antrum, which results in submucosal vein obstruction and GIA; however, Kim et al reported that a non-antral (gastric corpus/fundus) location was a risk factor for active bleeding in UGIA. Additionally, vascular endothelial growth factor (VEGF) has been recognized as the root cause of aberrant angiogenesis in UGIA, and researchers hypothesized that the gastric corpus may be injured by endogenous stimuli caused by high levels of VEGF.[13,29–31] More definitive evidence is needed to determine the endoscopic predictors that influence active bleeding in UGIA. Clinicians must also determine whether endoscopic hemostasis is required in select patients. The active bleeding in GIA may be occult or life-threatening, and active bleeding episodes often require treatment with fasting, volume repletion, blood transfusion, somatostatin analogs, hemostatic agents, and endoscopic intervention. At present, the endoscopic interventions recommended for bleeding in GIA include APC, thermal contact devices, such as monopolar and bipolar electrocautery, heater probes, hemostatic graspers, hemoclipping, cryotherapy, band ligation, and sclerotherapy. If bleeding persists following endoscopic intervention, percutaneous embolization or surgery should be considered.[32,33] In our study, most of the patients with active bleeding (71.4%) underwent endoscopic hemostasis with APC or the hemostatic grasper. The rebleeding rate was higher in the active bleeding group than the asymptomatic group (21.4% vs 0%), and the rebleeding rate in this study was similar to the rates published in previous studies (27% to 42%).[13,34] Grooteman et al identified liver cirrhosis (OR, 4.0), CKD (OR, 4.5), prothrombin time < 30% (OR, 4.2), multiple lesions (OR, 4.2), and location in the third small bowel (OR, 4.3) as risk factors for rebleeding.[35] We attempted to identify factors that influenced recurrent bleeding in this study, but none of clinical or endoscopic factors achieve statistical significance. Further better-designed research is warranted to verify our findings. Our study has some limitations. First, it was a single-center retrospective study that examined a small sample size. As such, the findings in this study may have been underpowered or subject to bias. Second, some of the EGDs in our cohort were performed by fellows in training, and some UGIA lesions may have been missed by unexperienced operators. Third, our database was comprised of Asian patients alone. Our study may have missed racial factors that predispose a group of patients to bleeding in UGIA, which reduces the external validity of our analysis. Despite these limitations, our study provided excellent data, because we analyzed an integrated endoscopic databank. In addition, we examined definitive cases of active bleeding in UGIA, because we diagnosed active bleeding through endoscopy instead of through clinical symptoms of GI bleeding or positive fecal occult blood test results. 5. Conclusions The epidemiology of UGI bleeding is changing. Despite the increasing incidence of UGIA, there are limited data on the risk factors that predispose patients with UGIA to active bleeding. This is the first cohort study to investigate the predictors of active and recurrent bleeding among patients with UGIA and provide real-world data on the clinical features and endoscopic presentation. The main point of this study revealed DM, CKD, and lesions in the duodenum were independent predictors of active bleeding in UGIA. Our study can help clinicians determine whether treatment of UGIA is warranted during endoscopy and further large-scale prospective randomized trials are warranted. Author contributions Conceptualization: Meng-Chuan Lu, Hsuan-Hwai Lin. Data curation: Hsuan-Wei Chen, Chih-Wei Yang, Jung-Chun Lin. Formal analysis: Tien-Yu Huang, Peng-Jen Chen, Wei-Chen Huang, Hsuan-Hwai Lin. Methodology: Tien-Yu Huang, Peng-Jen Chen, Wei-Chen Huang, Hsuan-Hwai Lin. Resources: Hsuan-Wei Chen, Chih-Wei Yang, Jung-Chun Lin. Software: Bao-Chung Chen. Supervision: Yu-Lueng Shih, Tsai-Yuan Hsieh. Validation: Yu-Lueng Shih, Tsai-Yuan Hsieh. Visualization: Bao-Chung Chen. Writing – original draft: Meng-Chuan Lu, Bao-Chung Chen. Writing – review & editing: Meng-Chuan Lu, Bao-Chung Chen.