Diagnostic and Therapeutic Yield of Endoscopy in Patients with Elevated INR and Gastrointestinal Bleeding

Gastrointestinal bleeding is a well-known risk of systemic anticoagulation. However, bleeding in the setting of supratherapeutic anticoagulation may have a milder natural history than unprovoked bleeding. It is a common clinical gestalt that endoscopy is common, but bleeding source identification or intervention is uncommon, yet few data exist to inform this clinical impression. Consequently, we sought to examine our institutional experience with gastrointestinal bleeding in the setting of supratherapeutic international normalized ratio (INR) with the aim of identifying predictors of endoscopically identifiable lesions, interventions, and outcomes.

A retrospective review was conducted at a tertiary referral academic medical center to identify patients presenting with gastrointestinal bleeding in the setting of warfarin and a supratherapeutic INR (>3.5) who underwent an endoscopic procedure. Relevant clinical covariates, endoscopic findings, need for intervention, and outcomes were collected by review of the medical record. Logistic regression adjusting for potential confounders identified predictors of endoscopically significant lesions as well as intervention and outcomes.

A total of 134 patients with INR 3.5 or greater (mean 5.5, range 3.5-17.1) presented with symptoms of gastrointestinal bleeding, most commonly as melena or symptomatic anemia. Antiplatelet agents were used by 54% of patients, and 60% of patients were on concomitant acid suppression on admission. Procedures included esophagogastroduodenoscopy (upper endoscopy; EGD) (n = 128), colonoscopy (n = 73), and video capsule endoscopy (n = 32). Active bleeding at first EGD or colonoscopy was found in only 19 patients (18%), with endoscopic intervention in only 26 patients (25%). At a critical threshold of INR 7.5 at presentation, the likelihood of finding an endoscopically significant lesion fell to <20%. On multivariate logistic regression, concomitant antiplatelet therapy (odds ratio [OR] 2.59; 95% confidence interval [CI], 1.13-5.94), timing of EGD within 12 hours of presentation (OR 3.71; 95% CI, 1.05-13.08), and INR level (OR 0.79; 95% CI, 0.64-0.98) were the only significant independent predictors of identifying a source of bleeding. A risk score incorporating these covariates performed modestly in identifying risk of significant finding on EGD (area under the curve 0.68). We found no association between identification of a significant lesion at EGD and future readmission for gastrointestinal bleeding.

This study demonstrates that the relationship between INR elevation and identification of a bleeding source or endoscopic intervention at EGD are indeed antiparallel. Concomitant antiplatelet therapy increases the likelihood of bleeding source identification and intervention, as does EGD within 12 hours of presentation. However, regardless of source identification or endoscopic intervention, important clinical outcomes were unchanged, suggesting that decisions about endoscopy should be made on a case-by-case basis, particularly in patients with INR > 7.5. Future prospective studies on appropriate indications and timing of endoscopy in such patients are warranted.