LAARGE is a prospective, non-randomised, multicentre real-world registry that encompasses patients from 37 voluntary participating centres. Its main objective is to represent the LAAC procedure’s clinical reality. For this reason, the protocol neither influenced indication nor clinical management, but it claimed consecutive enrolment to avoid a recruitment bias. Devices should be implanted according to current recommendations. Recruitment in the registry started in July 2014 and ended in January 2016.

For the present subanalysis, patients with started procedure and documented renal function were selected from the whole database. The study was carried out according to the principles of the declaration of Helsinki and was approved by the ethics committee of the State Chamber of Medicine in Rhineland-Palatinate, Germany. Written informed consent was obtained from all study patients.

The eGFR was calculated using The Modification of Diet in Renal Disease (MDRD) Study formula [8]. According to the guidelines, patients with eGFR < 60 mL/min/1.73 m² were categorised as having a clinical relevant impaired renal function [13]. Patients with CKD were categorised into three groups (i.e. eGFR < 15 mL/min/1.73 m², eGFR 15–29 mL/min/1.73 m², and 30–59 mL/min/1.73 m²).

As described previously [14], the preprocedural screening, the conduction of the implantation procedure as well as the postprocedural management including the antithrombotic treatment were at the discretion of the operating physician. Different standard commercial devices were implanted taking into consideration the specific manufacturer’s recommendations.

All participating centres reported procedural data and intra-hospital complications as well as discharge medication via an electronic case report form. Patients were contacted directly or via phone call one year after the implantation procedure to assess the survival, the occurrence of complications, and the antithrombotic treatment. If no contact could be established with a patient, information was obtained from the registration offices. For the purpose of data validation, all relevant events were reviewed and evaluated by an Endpoint Adjudication Committee, if necessary based on the original medical documents.

The effectiveness was primarily assessed by the absence of all-cause death and stroke during follow-up, secondarily by the absence of transient ischemic attacks (TIA) and systemic embolism. The implantation success was defined as a stable device anchorage. Complications including para-device leaks > 5 mm, device dislocations, severe and moderate bleedings during hospitalisation and during follow-up as well as thromboembolism in the venous system represented the safety outcome measure.

Statistical analyses were performed with SAS^® version 9.4 (SAS Institute, Cary, NC, USA). Continuous data are presented as means with standard deviation or as medians with interquartile ranges (25th and 75th percentiles), categorical data as frequencies with group-related percentages. Trends across the patient groups were assessed by a Cochran-Armitage test regarding categorical variables, or by an exact Cochran-Armitage test in case of rare events, and by a Jonckheere–Terpstra test regarding metrical variables, as indicated in the tables. In addition, CKD patients were compared to non-CKD patients using the Pearson Chi-squared test or Mann–Whitney–Wilcoxon test for categorical and metrical variables, respectively. These statistics were based on the available cases.

The one-year mortality after the implantation procedure and the incidence of the combined event of death or stroke were evaluated by methods of survival analysis (Kaplan–Meier curves, log-rank test). Hazard ratios with 95% confidence intervals (CI) were estimated using Cox regression without adjustment and adjusted for baseline characteristics significantly associated with CKD and known as clinically relevant risk factors: age (linear), sex, body mass index > 25 kg/m², arterial hypertension, diabetes mellitus, coronary artery disease, congestive heart failure, and LVEF ≤ 40%. The expected annual rates of major bleeding and stroke were calculated from the individual HAS-BLED [9] and CHA₂DS₂-VASc score, respectively [15]. The follow-up duration was defined as the time span from the index discharge to the date of the follow-up contact. p values ≤ 0.05 (two-tailed) were considered significant.

623 patients were included in the present analysis. 299 (48.0%) revealed a CKD (Table 1). The median eGFR value was calculated at 41.1 vs. 78.8 mL/min/1.73 m³ (p < 0.001 for the comparison to non-CKD patients). CKD patients were significantly older (77.8 ± 7.5 vs. 74.4 ± 7.8 years, p < 0.001) and revealed a significantly higher stroke (CHA₂DS₂-VASc score 4.9 ± 1.5 vs. 4.2 ± 1.5, p < 0.001) and bleeding risk (HAS-BLED score 4.3 ± 1.0 vs. 3.5 ± 1.0, p < 0.001), whereby an HAS-BLED score ≥ 3, corresponding to a high bleeding risk, was significantly more frequent in CKD patients (97.6 vs. 84.7%, p < 0.001). CKD patients also revealed a more pronounced cardiovascular risk profile.

Participating centres could document more than one indication for LAAC in the same patient (Table 1). Across all predefined GFR groups, the main indication was a prior bleeding event (79.8%; p = 0.022 for trend).

Supplemental Table 1 shows data from cardiac imaging procedures. While left atrial diameters were larger in CKD patients (p = 0.024 for trend), this finding did not correspond with the LAA diameters (each p > 0.05 for trend).

Technical success was high across all groups (97.9%; p = 0.76 for trend; supplemental Table 2), and no peri-device leak > 5 mm was present. Three interventions had to be interrupted prematurely (p = 0.87 for trend). A stable device anchorage could not be achieved in additional three patients. Device selection and dimensions as well as procedural parameters did not differ significantly (each p > 0.05 for trend).

Intra-hospital complications, and particularly major adverse and cerebrovascular events (MACCE) were generally rare (each p > 0.05 for trend; Table 2). Correspondingly, time to discharge was generally short (p = 0.097 for trend). Two intra-hospital deaths among the CKD patients were due to either an unknown or of cardiovascular aetiology, respectively. Seven dislodged devices could be retrieved catheter-based (each p > 0.05 for trend). Antithrombotic discharge medication did not differ significantly (each p > 0.05 for trend; supplemental Fig. 1), provided that 12.2% of patients stayed on anticoagulation when leaving the hospital (p = 0.57 for trend).

A total of 608 patients (97.9%) could be followed-up (p = 0.85 for trend; Table 3). Limited to 365 days after the procedure, the combined primary effectiveness outcome measure was reached in 82.0% among CKD patients and 93.0% among patients without impaired renal function (p < 0.001; Fig. 1). Even after adjustment for relevant risk factors, this effect was still present (Fig. 2), but there was no statistically significant difference when comparing the patient groups with an eGFR < 60 mL/min/1.73 m² among each other (p = 0.76). Only three non-fatal strokes were observed in the total cohort (p = 0.25 for trend), which all were ischemic. Moreover, rates of TIA and systemic embolism were low cross all GFR groups (each p > 0.05 for trend). Severe (p = 0.021 for trend) and moderate bleedings (p = 0.52 for trend) were infrequent across all groups. Despite only 6.0% of patients received anticoagulation after one year (p =  0.13 for trend), only two deep vein thromboses were registered (p = 1.00 for trend). 89.6% of patients were completely content with the intervention, and 96.6% of patients felt safe during hospital stay (each p > 0.05 for trend).

These analyses were based on observational registry data with the inherent limitations of this study type. The conduction of the intervention was not influenced by the study protocol and based on the operators’ discretions as well as the relevant recommendations, which respected the observational character of the registry. This individualised decision algorithm may have had impact on the outcome measures but surely reflects the clinical practice. The implantation volume per centre and per operator was naturally heterogeneous, which also meant a good mixture of experience. Though a separate group with renal replacement therapy was envisaged, there were not enough cases to perform an individual analysis on such patients. Despite these limitations, this registry is surely serving as a data source for a little-studied topic.