Impaired kidney function at ED admission: a comparison of bleeding complications of patients with different oral anticoagulants

This is a retrospective cohort study of the adult ED of the Bern University Hospital, Inselspital, Switzerland. Our ED is responsible for the emergency treatment of about 50,000 patients per year from a catchment area of 2 million people.

Primary study outcomes were bleeding and patient characteristics on DOAC therapy compared to VKA.

Secondary outcomes were procedural outcomes (length of stay in the ED and in hospital, ICU transmission and hospitalisation rate as well as ED resource consumption) and clinical outcomes, i.e. in-hospital mortality.

The National Kidney Foundation Kidney Disease Outcomes Quality Initiative (KDOQI) classification was used for classification of kidney impairment [15], with the severity grades according to the GFR of 45–60 mL/min, 30-44 mL/min, 15-29 mL/min, and < 15 mL/min. The GFR was calculated using the CKD-EPI equation as it was recommended by the National Kidney Foundation because of the reliability across all CKD stages [16]. Although the Cockroft-Gault equation for GFR estimation was applied in the landmark clinical trials of DOAC in atrial fibrillation, it has several limitation in its application particularly in patients that are overweight and older [7].. Since the implementation of the Cockroft-Gault equation was not feasible in our retrospective ED study as patient weight is not documented for all ED patients we performed the analysis using the CKD-EPI equation which is calculated automated in all patients. Polypharmacy was defined as the use of five or more drugs [17].

To describe comorbidities of a patient, the Charlson comorbidity index was used including parameter such as liver disease, CKD, diabetes, cerebrovascular disease, and congestive heart failure [18].

Major bleeding were defined for our study according to the international society of thrombosis and haemostasis (ISTH) definition as fatal bleeding, symptomatic bleeding in critical organ or transfusion of at least two erythrocytes concentrates [19]. As a documented haemoglobin drop of at least 20 g/L was not feasible to determine reliable in the ED setting, the criterion was not used.

In Switzerland, the costs incurred in an ED for each ambulatory procedure or physician/nurse work are settled in “tarmed: tarif médical” points which, is a system of procedure codes. This point system used in Switzerland corresponds to 1 Swiss franc per 1 point billed. We indicate the total ED resource consumption during treatment in “tax points” as the sum of all procedures.

A keyword search was performed through all medical reports between 01.06.2012 to 01.07.2017 containing all the substance classes and brand names of oral anticoagulants approved in Switzerland (substance classes: rivaroxaban, dabigatran, apixaban, and edoxaban, phenprocoumon, warfarin, and acenocoumarol), combined with the logic operator “OR”. Medical ED reports in our department including patient history, diagnosis, clinical findings, and medication administered are stored electronically in an ED patient database (E-Care, ED 2.1.3.0, Turnhout, Belgium). We excluded all search results of cases without creatinine testing at ED admission. Finally, the eligibly criteria were evaluated by manual full text analysis of each found medical ED report.

The following parameter were extracted directly from the database: age, sex, triage level (which is routinely determined by special trained nurses using the Swiss triage scale [20]), type of referral, all procedural outcomes, in-hospital mortality, as well as creatinine and INR values on admission. Using full-text analysis of the medical report, the following parameter were manually extracted: trauma, information on anticoagulation (indication for therapy, type of medication, additional antiplatelet therapy), comorbidities to calculate the Charlson comorbidity index, arterial hypertension and polypharmacy, bleeding characteristics i.e. location and extent of bleeding (major bleeding and the discharge diagnosis of a haemorrhage shock) as well as information on the therapy of the bleeding. Interventional therapies included surgery, coiling, clipping, embolization, angiography, drainage, and variceal ligation. A diagnostic endoscopy was not considered an interventional therapy.

The DOAC group used for analysis combined all patients on any DOAC approved in Switzerland (apixaban, dabigatran, edoxaban and rivaroxaban) on ED admission. For group comparison we formed a VKA group, including all patients under phenprocoumon and acenocumarol at ED admission. All statistical analysis for this work was performed with Stata® 13.1 (StataCorp, The College Station, Texas, USA).

For descriptive analysis the distribution of categorical variables was presented as the absolute accompanied by the relative number in each category [n (%)] and of continuous outcomes with the median accompanied by the interquartile range (IQR) as normal distribution of continuous outcomes could not be ensured.

The present study is registered with the competent ethics committee of Canton Bern, Switzerland under the number 073/2015 and according to Swiss law, the need to obtain informed consent for the study was waived.

An overview of all patient characteristics is presented in Table 1.

There were no significant differences regarding gender and age (p = 0.771 and p = 0.774) between both anticoagulation groups (see Table 1). The renal function in patients on VKA was significantly (p = 0.002) worse compared to patients on DOAC. Especially in the GFR < 15 mL/min group 7.4% patients were on VKA vs. 0.6% on DOAC. Significant more patients (p = 0.029) on DOAC (5.7%) had a documented liver insufficiency compared to VKA (2.1%). The Charlson comorbidity index (CCI) had a median of 5 (IQR 4–7) points in both groups (p = 0.143). There was no significant group difference regarding polypharmacy (p = 0.352).

The type of referral or triage category was similar in both two study groups (Table 2). Trauma was found more often in VKA patients (32.9%, n = 138) compared to DOAC patients (24.1%, n = 38, p = 0.039). In the DOAC group, the majority of DOAC patients were on rivaroxaban (n = 141, 89.2%). Nine patients were on apixaban (5.7%), six on dabigatran (3.8%), and two (1.3%) on edoxaban. In the VKA group, the majority of 415 patients were on phenprocoumon (99.0%); 4 patients were on acenocumarol (1.0%). Regarding indication for anticoagulation, the most common indication for anticoagulation in all patients was atrial fibrillation (49.7%). Significant group differences were found regarding “mechanical heart valve” and “venous embolism” as indications for anticoagulation (p = 0.009). Additional antiplatelet therapy did not differ significantly between the two study groups (p = 0.346).

The median INR in patients with VKA was 2.4 (IQR: 1.7–3.3), and in patients with DOAC 1.3 (IQR 1.1.-1.5). Using the INR range 2.0–3.0 to the definition of therapeutic range, 31.3% of the VKA patients were over and 34.4% were under the therapeutic range.

Apart from a higher number of intracranial bleedings including intracerebral bleedings in the VKA group (22.4% vs. 14.6% in DOAC patients, p = 0.036), no significant differences in the distribution of bleeding locations were found, see Table 3. The most common location was GIB in both anticoagulation groups (DOAC 34.8% vs. VKA 28.6%, p = 0.135). The incidence of major bleeding events did not differ significantly between the study groups (p = 0.784). Major bleeding was not significantly associated with the GFR in the DOAC (p = 0.723) respectively VKA group (p = 0.795).

VKA patients needed a higher number of interventions to stop the bleeding compared to DOAC (22.2% vs. 13.9%, p = 0.027) and a smaller number needed gastroduodenoscopies (VKA 9.1% vs. DOAC 15.8%, p = 0.020), see Table 4.

Most patients were treated in hospital (DOAC 92.4% vs. VKA 90.2%, p = 0.431) and 42.0% were admitted to ICU (DOAC 38.2% vs. VKA 43.4%, p = 0.264) without group difference for both outcomes, see Table 5. No differences were observed regarding ED resource consumption, LOS in hospital or the ED as well as in-hospital mortality.

Most of the bleeding locations occurred in a comparable frequency in both, DOAC and VKA patients with impaired renal function. GIB are a typical bleeding location for DOAC especially on rivaroxaban therapy [24]. For both anticoagulant groups, GIB were the most common bleeding leading to ED admission in our study. Intracranial bleedings were more prevalent in the VKA group. As the study population consisted of patients admitted for major hemorrhage, less frequent occurrence of one bleeding location might be caused by a higher frequency of another bleeding location and vice versa. Thus, only proportional incidences could be determined in this study.

Patients with GIB under DOAC therapy received more endoscopies overall and especially more gastroscopies compared to VKA, although VKA patients received more interventional therapies. Several reasons might explain these findings. First, physicians’ management of DOAC-associated bleedings might differ as the specific reversal agents of DOACs are expensive and may not be available in some ED. Therefore, treating physicians are more cautious as a multi-nominal survey about bleeding management demonstrated [25]. This fact is reflected by another study of our group that found an increased application of prothrombin complex concentrates for reversal of DOAC compared to VKA bleedings in clinical practice [26]. A recent meta-analysis of PCC and andexanet alfa for management of factor Xa inhibitor related bleeding, found similar rates of good hemostasis for both medications but a tendency to higher complication rates on andexanet alfa [27]. Second, several studies suggested a higher rate of GIB in DOAC patients [5, 28], thus, a wait-and-see approach might seem an inappropriate choice for the attending physician in DOAC patients with suspected GIB as the extend of the bleeding in often unclear. Third, VKA reversal is straightforward und routinely measurable compared to DOAC, thus, some interventional procedures might have been performed in VKA patients, while DOAC interventions were postponed.

No significant differences were found in the investigated outcomes between the DOAC and VKA group. The hospitalisation and ICU transmission rates were high, a finding that is explained by the high number of GIB and intracranial bleedings in both groups needing close clinical monitoring or further invasive therapies.

Due to the limitations imposed by the size of the study and the retrospective design, there are potential confounders that were not considered in our study. Furthermore, despite the adequate documentation of complications in a tertiary hospital setting and high number of patients on both anticoagulant group, it is possible that the group sizes were still not sufficient in order to find differences in infrequent outcomes such as in-hospital mortality. Because of the monocentric design transferability to other settings is only possible to a limited extent and should be performed carefully. These limitations are not unique to our study, but rather apply to most observational pharmacoepidemiologic studies. Investigation of real-word data can be an important complement to randomized-controlled-trails, however, conclusions must be drawn with great caution due to the uncontrolled confounding factors [13, 14]. Also, we had no information on the chronicity of CKD or on the duration of the anticoagulation therapy. A single creatinine obtained during emergency admission can only provide a snapshot of the patient’s renal function. The same applies for the INR values, e.g. after reversal of VKA with PCC. Furthermore, levels of DOAC or aPTT were not measured routinely over the study period, thus the decision how to treat a patient on DOAC was often based on anamnestic data of DOAC intake. There is growing evidence, that NOAC level measurements should be standard practice, especially in at-risk populations such as patients with renal insufficiency [29].

It is well known that decisions in emergency patients have to be based on often incomplete findings and have to be made under time pressure, which does not allow the detailed collection of follow-up parameters [30‐32]. In line with the local prevalence of anticoagulant medications most investigated patients were admitted on rivaroxaban [2]. Thus, separate analysis or comparison of different DOAC or stratified presentations of the outcome by GFR group was not possible due to small numbers of patients in those subgroups. With the fast adoption of DOAC including elderly patients over the study years, more DOACs were subscribed to patients with pre-existing conditions and therefore at risk for an impaired kidney function and bleedings. With our data, the exact impact on the results cannot be stated. However, in our opinion, the main findings of a lower proportional incidence of intracranial haemorrhage, more emergency endoscopies and less interventional therapies in DOAC patients remain untouched by this finding.