Left atrial appendage occlusion in patients with atrial fibrillation and major gastrointestinal bleeding: outcomes from a multi-hospital health system

While left atrial appendage occlusion (LAAO) has been shown to be non-inferior to warfarin for stroke prevention in patients with atrial fibrillation (AF), the trials leading to Food and Drug Administration approval have excluded patients with a previous history of major gastrointestinal (GI) bleeding (MGIB).1 Conversely, a history of MGIB is one of the most frequent clinical indications for LAAO, but no randomized controlled trial has analysed LAAO specifically in this patient population. We therefore aimed to compare outcomes of LAAO in patients with previous MGIB to a propensity-matched control group with previous MGIB pursuing continued oral anticoagulation (OAC) in a real-world setting. Consecutive patients implanted with LAAO device and previous MGIB at the hospitals of the University of Pittsburgh Medical Center from January 2016 to November 2022 were included (n = 633). These patients were propensity matched to patients with AF and previous MGIB based on 17 baseline characteristics (Figure 1). The beginning of follow-up was defined as the time of LAAO implantation, or as the MGIB event in the group who continued OAC. Major gastrointestinal bleeding was defined as upper and/or lower GI bleeding requiring inpatient hospitalization or necessitating blood transfusion. Left atrial appendage occlusion was performed as described previously.2 Only Watchman and Amplatzer devices were used and chosen according to operator’s preference as well as pre-procedural imaging. Post-procedural anticoagulation/antithrombotic therapy was managed according to current recommendations.3 Clinical outcomes including all-cause survival, all-cause admissions, cardiac admissions, and admissions for ischaemic stroke/transient ischaemic attack (TIA), haemorrhagic stroke, MGIB, and any major bleeding were assessed during clinical follow-up. Statistical analyses were performed using Stata (Stata Corp. 2025, College Station, TX). Since the groups were not balanced with regard to baseline OAC use, Cox regression models with additional adjustment including baseline OAC use were performed to calculate adjusted hazards ratios (aHRs). Statistical significance was assumed if P < 0.05. All research activities were approved by the University of Pittsburgh internal review board. Outcomes of LAAO in patients with previous MGIB. Kaplan–Meier curves demonstrating freedom from MGIB (A), any major bleeding (B), ischaemic stroke/TIA (C), and all-cause survival (D). (E) Adjusted hazard ratios following adjustment for baseline use of oral anticoagulation. Propensity score matching was performed using 17 variables including age, body mass index, race, gender, history of hypertension, diabetes, chronic obstructive pulmonary disease, coronary artery disease/peripheral vascular disease, previous coronary artery intervention, liver cirrhosis, chronic kidney disease, cancer, ischaemic stroke, haemorrhagic stroke, use of non-steroidal anti-inflammatory drugs, use of warfarin, and HAS-BLED and Elixhauser comorbidity scores. Values in panel (E) are presented as frequency (percentage). Adjusted hazard ratios are presented with (95% confidence intervals). aHR, adjusted hazard ratio. LAAO, left atrial appendage occlusion; MGIB, major gastrointestinal bleeding; TIA, transient ischaemic attack. Both groups were well balanced at baseline (n = 1,266, mean age 76.8 ± 8.9 years, mean CHA2DS2-VASc score 4.3 ± 1.7, mean HAS-BLED score 3 ± 0.8). All major comorbidities including chronic obstructive pulmonary disease, hypertension, coronary artery disease/peripheral vascular disease, ischaemic and haemorrhagic strokes, cirrhosis, chronic kidney disease, and cancer were similar in both groups (all P > 0.05). At baseline, fewer patients in the LAAO group were on OAC (59.7% vs. 78.2%; P < 0.001), with fewer patients being on dabigatran (3.6% vs. 10.3%, P < 0.001) and rivaroxaban (19.3% vs. 28.9%, P < 0.001). Over a median follow-up of 30.9 months (interquartile range 17.1–49.1 months), patients in the LAAO group had fewer all-cause admissions (365.9 vs. 525.8/1000 person-years, P < 0.001), cardiac admissions (253.4 vs. 341.4/1000 person-years, P < 0.001), admissions for ischaemic stroke/TIA (15.6 vs. 25.5/1000 person-years, P = 0.045), MGIB (97.2 vs. 129.7/1000 person-years, P = 0.009), any major bleeding including procedure-related bleeding (97.2 vs. 130.9/1000 person-years, P = 0.007), and lower all-cause mortality (88.9 vs. 121/1000 person-years, P = 0.003) compared with patients who continued oral anticoagulation (Figure 1A–D). These results persisted after adjustment for baseline OAC use (all HRs with P < 0.05; Figure 1E). Admissions for haemorrhagic stroke, on the other hand, did not differ between the groups (6.8 vs. 9.4/1000 person-years, P = 0.39), with comparable adjusted hazards (HR 0.65, P = 0.28). To our knowledge, we present the first study to compare outcomes of LAAO in patients with previous MGIB compared with continued OAC. Our study revealed that, in comparison to OAC, LAAO was associated with a reduction of recurrent MGIB and all-cause bleeding events, lower rates of ischaemic stroke and TIA, and improved all-cause survival. PROTECT AF and PREVAIL, the landmark trials comparing LAAO to OAC with warfarin, excluded patients with previous MGIB.1 While the most recently published OPTION trial demonstrated a reduction in non–procedure-related major bleeding in patients who underwent AF ablation with concomitant LAAO compared with continued OAC, it is unclear how many patients had a previous MGIB, with the mean HAS-BLED scores being only 1.2 in both groups.4 Conversely, preliminary results of the CLOSURE-AF trial indicate inferiority of LAAO for its composite outcome compared to best medical care in high-risk patients, though not all included patients had a previous bleeding event.5 The only available studies on LAAO specifically in patients with previous MGIB were observational and compared major bleeding events to predicted event rates derived from the HAS-BLED score, thereby lacking a validated control group.6,7 Our study, on the other hand, included a propensity-matched control group and demonstrated a robust reduction of both MGIB and other major bleeding events in patients receiving LAAO when compared to OAC. Moreover, our results indicated improved cardiovascular and all-cause survival with LAAO in comparison to continued OAC. This is consistent with the results from the pooled 5-year analysis of the PROTECT AF and PREVAIL trials, which demonstrated improved cardiovascular and all-cause mortality in patients undergoing LAAO.1 Similarly, large-scaled, multi-centre registries have shown improved all-cause survival in patients undergoing concomitant surgical LAAO during cardiac surgery, though antithrombotic strategies were not consistent among treatment arms.8,9 Our findings suggest that, even in patients with previous MGIB, LAAO might confer a survival benefit when compared to continued OAC, even if survival post-LAAO is conditioned in the real world by a substantial risk of non-cardiovascular deaths.10 Several limitations need to be considered when interpreting our findings. First, our comparison is non-randomized and therefore susceptible to selection bias. Specifically, patients with a high predicted chance of long-term survival after successful LAAO procedure might have been preferentially selected for LAAO. Secondly, our analysis was limited to a single control group which continued OAC and does not allow for conclusions regarding different comparator groups, such as patients treated with antiplatelet therapy only or patients not receiving any form of antithrombotic/anticoagulatory regimen. Thirdly, admission rates served as a surrogate for clinical endpoints, and clinical events that did not lead to hospital admission are hence not accounted for. Finally, while our data were obtained from a single hospital system, it includes a wide range of community and tertiary care centres. In conclusion, this large, real-world analysis indicates superior outcomes, including fewer major bleeding and stroke events with improved survival in patients with a history of MGIB undergoing LAAO when compared to continued OAC. Our findings are impactful to clinical practice and worthy of further investigation in a prospective manner. Grant support from the Boston Scientific Corporation was received as part of an investigator-initiated project. The data underlying this article will be shared on reasonable request to the corresponding author.