FormalPara Key Points

A significant proportion of non-vitamin K antagonist oral anticoagulants are inappropriately dosed, which compromises its efficacy in stroke prevention in patients with atrial fibrillation.

Dosing of these oral anticoagulants can be based on different dosing recommendations.

Insights into deviating dosing decisions can improve real-life stroke prevention, a cornerstone of atrial fibrillation management.

1 Introduction

Non-vitamin K antagonist oral anticoagulants (NOACs) are now the standard of care for stroke prevention worldwide in patients with atrial fibrillation (AF) with high thrombo-embolic risk and in the absence of mechanical prosthetic heart valves or moderate/severe mitral stenosis, or severely depressed renal function [1, 2]. Currently, four NOACs are available in Europe, each with specific dose reduction criteria defined in their respective ‘Summary of Product Characteristics (SmPC) documents’. These criteria include age, renal function, weight and specific concomitant intake of medication. However, in daily practice, several less researched relevant aspects can influence the decision of clinicians to prescribe a different dose than recommended by the SmPC. This is why the European Heart Rhythm Association (EHRA) has developed sequentially updated practical guides for healthcare professionals concerning the use of NOACs in AF patients incorporating the SmPC criteria and important patient characteristics (e.g., frailty, concomitant use of antiplatelets) to provide support and scientific evidence concerning the dosing and use of NOACs [3,4,5,6]. Nevertheless, real-world studies have shown that a significant portion of AF patients treated with a NOAC receive inappropriate NOAC doses. Underdosing can lead to a higher risk of stroke, and overdosing can impair safety outcomes of these oral anticoagulants [7].

2 Aims

The primary aim of this retrospective study was to investigate whether there is a difference in the perceived appropriateness of NOAC dosing with respect to the SmPC or the 2018 EHRA Practical Guide in AF patients presenting at an outpatient visit at the cardiology department of the Antwerp University Hospital or at six primary care centres (all located in the Antwerp region).

The secondary aims were (i) to explore if there was a significant difference in appropriateness of NOAC dosing between primary care and specialist care, and (ii) when renal function was calculated according to different formulas.

3 Ethics Approval

The research protocol was approved by the Ethics Committees of the Antwerp University Hospital/University of Antwerp on the 12th of August 2019 and the study was conducted in compliance with the Declaration of Helsinki (local project reference 19/27/331).

4 Methods

4.1 Study Population and Enrolment Procedure

Patients were eligible for this study if they were (i) aged ≥ 18 years, (ii) diagnosed with AF or atrial flutter on an electrocardiogram and (iii) chronically treated with one of the four NOACs, namely apixaban, rivaroxaban, edoxaban or dabigatran. The indication of whether a patient should be treated with a NOAC was checked based on his/her CHA2DS2-VASc score. For the eligible patients of the Antwerp University Hospital, no explicit informed consent (IC) was needed as data were internally available and retrospectively retrieved by the study investigators; all patients of the hospital consented with inclusion in retrospective analyses. Enrolment of AF patients at the primary care centres was done consecutively by the general practitioner (GP), who explained the study to the patient and obtained the IC. Atrial fibrillation patients already enrolled in an interventional NOAC study or patients who were unable to sign the IC (i.e., language barrier) were excluded.

4.2 Data Collection

After approval of the research protocol, patients who had presented to any of the outpatient clinics after April 2018 (the date of publication of the EHRA 2018 Practical Guide) were retrospectively screened for inclusion. The inclusions were performed consecutively and equally spread over four cardiology subspeciality clinics (Interventional, Electrophysiology, Heart Failure and General Cardiology/Cardiac Imaging) to ensure a homogeneous AF cohort in follow-up by cardiologists. If a patient was found to have multiple visits, only the first clinic visit was assessed.

As mentioned previously, primary care patients were enrolled only after consent had been given. Recruitment was performed in AF patients presenting after approval of the research protocol (August 2019). Then, the medical file was reviewed retrospectively and the patients’ first GP visit after April 2018 was assessed for data extraction in order to have similar time periods evaluated in cardiology and GP patients.

Patients’ medical data were retrieved from the electronic patient record and included age, sex, actual body weight, height, body mass index (BMI), blood pressure, type of AF, prescribed NOAC and dose, concomitant medication and serum creatinine closest to the index consultation. Patients’ medical history was checked for components of the CHA2DS2-VASc and HAS-BLED scores. Moreover, a history of gastrointestinal bleeding, bleeding predisposition, recent surgery on a critical organ and available data to estimate frailty, based on the parameters used in the ENGAGE-AF TIMI 48 trial, were recorded as these factors also play a role in the 2018 EHRA Practical Guide [8].

Based on the collected data, calculation of renal function was estimated using the Cockcroft and Gault-(CG), the Modification of Diet in Renal Disease- (MDRD) and the Chronic Kidney Disease Epidemiology Collaboration- (CKD-EPI) equations [9,10,11].

Non-vitamin K antagonist oral anticoagulants dosing was evaluated by comparing the actual prescribed dose with the recommendations from the SmPC (Supplementary Table 1) and EHRA 2018 Practical Guide [12,13,14,15]. Classification was either appropriate or inappropriate in case of underdosing or overdosing. The EHRA 2018 Practical Guide incorporates additional clinical parameters that may justify dose adjustments and also includes an extensive list of interacting drugs that are not all included in the SmPCs (e.g., extended list of interactions with anticancer and antiepileptic drugs) [5]. This guide also uses a colour code with one important guide rule that recommends consideration of dose adjustment or the use of a different NOAC with fewer interactions (if available) in the presence of ≥ 2 ‘yellow’ criteria. Consequently, the EHRA 2018 guide is less stringent in case of a combination of ‘yellow’ criteria, which the SmPC dose adjustment criteria do not consider. For example, a 77-year-old patient with concomitant use of antiplatelets and a standard dose NOAC was classified as ‘appropriate’ for both classification systems (‘75+’ and ‘concomitant antiplatelet drugs’ are both yellow factors). The same patient on a reduced NOAC dose would be classified as ‘inappropriate’ according to the SmPC, but potentially ‘appropriate’ according to the EHRA 2018 Practical Guide.

Other principal colour codes include: ‘Orange’ = consider dose adjustment or different NOAC; ‘Red’ = contraindicated/not recommended; ‘Brown (dark)’ = contraindicated due to reduced NOAC plasma concentrations.

4.3 Sample Size

For the primary objective, a sample size of 152 AF patients in each arm was calculated using an alpha of 0.05 and a power of 0.80. This was based on a 10.4% difference in NOAC dosing appropriateness derived from a retrospective cohort study that investigated the correct prescription of NOACs in hospitalised patients comparing the SmPC prescription rules and the 2015 EHRA guide [16]. For the second objective of specialist care versus primary care, a sample size of 171 AF patients in each arm was calculated (alpha = 0.05 and power = 0.80), based on a substudy of the ORBIT AF-II registry, which reported data of incorrect NOAC dosing by different medical specialities (based on US approved package inserts) [17]. Combining these two sample size calculations, and anticipating 15% of incomplete patient files, an inclusion target of 197 AF patients, for both specialist and primary care was set forward (in total 394 patients). However, due to the COVID-19 pandemic, we did not reach the target inclusion rate in the GP cohort due to the severe impact on consenting procedures.

4.4 Statistics

Data were analysed using IBM SPSS version 27.0. Variables were described as numbers and percentages or as mean ± standard deviation, as appropriate. For continuous variables, differences between two (un)paired groups were compared using the paired-samples t test or independent-samples t test. The chi-squared test, the McNemar test and Fisher’s exact test were used for categorical variables, as appropriate. All comparisons were tested two-sided. P-values < 0.05 were considered statistically significant.

The relative risks (RRs) and odds ratios (ORs) were calculated and reported with their 95% confidence intervals (CIs) for significant categorical predictors for inappropriate dosing of NOACs (i.e., under- and overdosing). For continuous variables, univariate logistic regression models were used to calculate the ORs (with their 95% CI), and p-values were derived from the likelihood-ratio test. Candidate variables, categorical as well as continuous, with a p-value < 0.10 were considered for multivariate regression analysis and the optimal regression model was composed using a backward elimination strategy.

5 Results

5.1 Patient Characteristics

A total of 294 AF patients were included in this study, of which 200 (68.0%) patients were recruited at the cardiology outpatient clinic and only 94 patients (32.0%) at the GPs’ office (between September 2019 and February 2020) (Fig. 1).

Fig. 1
figure 1

Enrolment procedure. AF atrial fibrillation, NOAC non-vitamin K antagonist oral anticoagulant, PC primary care, UZA Antwerp University Hospital

Full size image

Table 1 presents the baseline characteristics of the included AF population. Mean serum creatinine was 1.09 ± 0.39 mg/dL for which the estimated renal functions calculated by the CG, MDRD and CKD-EPI formulae were 70.3 ± 28.8 mL/min, 71.2 ± 23.4 mL/min/1.73 m2 and 65.5 ± 20.3 mL/min/1.73 m2, respectively. Apixaban was the most commonly prescribed NOAC (41.5%) followed by rivaroxaban (34.4%), edoxaban (13.6%) and dabigatran (10.5%) with a reduced dose in 26.2%, 21.8%, 15.0% and 41.9% for each NOAC, respectively (p = 0.066; Table 2).

Table 1 Baseline characteristics of the study population
Full size table
Table 2 Appropriateness of NOAC dosing
Full size table

When comparing patients included in primary care versus specialist care, AF patients followed by GPs were older (78.6 ± 7.3 years). Consequently, they had a lower renal function calculated by the CG formula (65.1 ± 26.1 mL/min). These patients were also less known to have congestive heart failure (24.5% vs 37.5%) and took fewer antiplatelet drugs (3.2% vs 15.5%) (Table 1).

5.2 Appropriateness of NOAC Dosing

In general, according to the SmPC and EHRA 2018 guide, a rather high proportion of patients received an inappropriate dosage of NOAC, in 19.4% and 15.6% of patients (p = 0.003), respectively (Table 2). The significant difference was driven by a more lenient interpretation of potentially correctly underdosed NOACs by the EHRA 2018 (4.0%, p = 0.003). Translated in absolute numbers, of the 31 underdosed SmPC patients, 12 patients (38.7%) received a potentially correct NOAC dose according to the EHRA 2018 guide. These patients were more often classified as frail (RR = 5.46; 95% CI 1.85–16.06; p < 0.001) and more often used amiodarone (RR = 2.98; 95% CI 1.44–6.14; p = 0.022). Similar results were observed in overdosed patients when classified according to SmPC or EHRA 2018 (8.8%; n = 26).

Figure 2 shows dosing appropriateness per NOAC according to the SmPC guidelines and the 2018 EHRA Practical Guide.

Fig. 2
figure 2

Appropriateness of NOAC dosing according to the SmPC and EHRA 2018 guide. EHRA 2018 European Heart Rhythm Association 2018 Practical Guide, NOAC non-vitamin K antagonist oral anticoagulant, SmPC Summary of Product Characteristics documents

Full size image

5.3 Influencing Factors for Under- and Overdosing of NOACs

Inappropriate NOAC underdosing according to the SmPCs was univariately significantly related to the use of diuretics and to weight (or BMI) (all p < 0.05; Table 3), with borderline relations with perceived frailty and drug or alcohol use. In multivariate analysis, frailty and higher body weight were the only significant factors. Based on the EHRA 2018 Practical Guide only the use of drugs/alcohol and a higher BMI were correlated with an inappropriate reduced dose in both univariate and multivariate analysis (Table 3).

Table 3 Factors related to underdosing of NOACs
Full size table

For overdosed NOAC patients (both according to the SmPC and EHRA 2018 guide as identified patients were identical), primary care, permanent AF, older patients, not taking apixaban, lower weight (or lower BMI) and lower renal function were factors significantly correlated with a higher risk for overdosing (univariate analysis). In multivariate analysis, patients not on apixaban and with lower renal function were associated with inappropriate overdosing of their NOAC (Table 4).

Table 4 Factors related to overdosing of NOACs
Full size table

5.4 Primary Care Versus Specialist Care

Although the number of recruited patients in GP care was too low (caused by COVID-19 circumstances—see above), GP care versus cardiologist care was not retained in any multivariate analysis of factors related to underdosing or overdosing (Tables 3 and 4). Nevertheless, patients in GP care showed a higher rate of inappropriate dosing compared to cardiologists, which was non-significant based on the SmPCs (24.5% vs 17.0%; p = 0.131) but significant based on the EHRA 2018 guide (22.3% vs 12.5%; p = 0.03). This seems to be mainly the result of inappropriate overdosing (Table 4; p = 0.039 univariate p-value), which could be an indication that cardiologists consider more factors to reduce dose.

5.5 Influence of Different Renal Function Estimation Formulae

When comparing appropriateness of dosing based on renal function calculated by CG-, MDRD-, or CKD-EPI formulae, no significant differences were seen between these formulas, for either the SmPC- or for the EHRA 2018-based evaluation (Supplementary Table 2). On the other hand, the significant difference between the SmPC and EHRA 2018 Practical Guide as described with the CG formula in section 5.2 (p = 0.003) remained significant when reclassifying appropriateness using the MDRD and CKD-EPI formulae (with p-values of < 0.001 and 0.002, respectively) (Supplementary Table 3).

6 Discussion

This study in ambulatory AF patients found a high prevalence of inappropriate NOAC dosing (19.4%) according to the SmPC. When based on the EHRA 2018 Practical Guide, the proportion is significantly lower (15.6%) but still, 1 out of 7 AF patients, seem to be receiving an inappropriate dose of NOAC. The explanation,i.e. whether prescribers are incorrect, or whether prescribers have good reasons beyond the guidance to adapt the dose, remains a topic of study. We identified several factors associated with inappropriate NOAC dosing. Of note, reclassification of NOAC appropriateness based on the MDRD and CKD-EPI renal function estimation formulae (which are more readily available to clinicians than to the CG calculation) did not explain the difference in the classification of dosing.

6.1 Prevalence of NOAC Misdosing

As AF prevalence is expected to increase in the upcoming decades, optimal treatment of these patients is necessary to minimise AF complications and reduce the health burden, for both patients and for healthcare systems. A cornerstone of AF management is the prevention of stroke, for which NOAC treatment is the first-choice therapy [18,19,20,21]. Besides identifying and treating AF patients with a high risk of stroke, correct NOAC dosing is also of primordial importance to ensure efficacy and safety.

The range of ambulatory AF patients treated with an inappropriate dosage of NOACs in our study is in line with other large international investigations ranging between 12.8–31.1% [17, 22,23,24]. Two smaller Belgian studies by other centres in our country reported off-label dosing in 25.0% and 18.3% [16, 25]. Remarkably, in the aforementioned studies overdosing ranged between 3.4–7.8%; whereas in our study overdosing was slightly more prevalent in 8.8% of patients.

When applying the EHRA 2018 Practical Guide, an expected (but significant) decline of inappropriate dosing was found (− 3.8%) compared with the SmPC. This was driven by more lenient acceptance of reduced NOAC doses as potentially appropriate (from 10.5 to 6.8%). Moudallel et al reported NOAC underdosing in 17.4 versus 7.0% of patients according to the SmPC and EHRA 2015 Practical Guide, respectively, and is in line with our results regarding NOAC underdosing (6.1% were overdosed according to the SmPC but no data were reported concerning overdosing according to the EHRA 2015 guide) [16]. Two other European studies also evaluated NOAC dosing appropriateness according to the EHRA 2015 guide, but interpreted the presence of ≥ 2 ‘yellow’ interactions as an indication for a reduced dose. A retrospective subanalysis of the FANTASIIA Registry (a Spanish prospective, observational, multicentre study including adults with AF on anticoagulant evaluating the incidence of thrombo-embolic and bleeding events) found inappropriate doses in 32% of AF patients. More specifically, 15% were inappropriately overdosed and 17% were inappropriately underdosed (off-label dosing according to SmPC criteria was not reported) [26]. Capiau et al found an increase of inappropriate dosing from 18.3 to 23.4% according to the SmPC and EHRA 2015 guide, respectively, (for both systems 0.8% of NOACs were contra-indicated) [25]. Of the SmPC underdosed patients (9.8%), 21.9% were correctly dosed when classified by the 2015 EHRA guide, resulting in 7.6% underdosed patients. Therefore, the global increase of dose inappropriateness was related to the increase of overdosed patients according to the EHRA 2015 guide (from 7.8 to 15.0%). Of note, since this interpretation of “≥ 2 yellow factors” is suggested as a possibility in the EHRA Practical Guide, we considered a standard dose as a reduced dose ‘appropriate’ in such cases, which explains the overall lower prevalence of inappropriateness in our study.

Since the EHRA Practical Guide considers more factors with relevance for dosing, we anticipated that incorrect dosing would be less when judged by the Practical Guide than by the SmPC. It shows that in daily life, a large proportion of AF patients have a complex presentation. Nevertheless, even when evaluated by the EHRA Practical Guide, standard inappropriate dosing is prevalent. This could be explained in two ways. One is that physicians correctly take more clinical factors into consideration and hence, both the SmPC and EHRA PG still fall short of guide clinical practice. However, prior retrospective and observational data have clearly shown that dosing that deviates from recommendations is associated with increased risk of adverse events and even mortality [7, 27]. Therefore, the second explanation is that physicians still are falling short of making correct dosing decisions, which calls for more physician education to improve patient outcomes. This education could focus more specifically on some of the factors that our research has shown to be related with prescription errors. It also calls for better patient tailored (transmural) follow-up with frequent reassessment of NOAC dose to improve results.

6.2 Contributing Factors for Under- and Overdosing of NOACs

Prior studies have identified various univariate factors related to inappropriately reduced dosing, such as age, CHA2DS2-VASc score < 4, sex (female), ethnicity (non-Caucasian), acute coronary syndrome, vascular disease, prior stroke, diabetes and concomitant antiplatelet therapy [23, 24]. The FANTASIIA Registry found that the factors ‘younger age’ and ‘dabigatran use’ were also associated with inappropriately low NOAC dosing [26]. Our study, also retained the (univariate) association of drugs (i.e., antiplatelets or NSAID) and alcohol with underdosing. Of note, alcohol is a factor in the HAS-BLED score and not in the SmPC or EHRA Guide, and is a modifiable bleeding risk factor that should be addressed rather than lead to an adaption of the NOAC dose.

Based on the SmPC multivariate analysis, the factors ‘higher body weight’ and ‘frailty’ were associated with off-label underdosing.

For overweight or obese patients, this is a paradoxical finding: although higher weight is associated with both a higher volume of distribution and higher renal clearance, no specific (i.e., increased) NOAC dosing algorithm currently exists. At least the standard NOAC dose would be expected. This suggests that other factors that are not even part of the SmPC or EHRA Guide led physicians to paradoxically reduce the dose. One could postulate that some conditions for dose reduction are more prevalent in overweight patients (e.g., vascular disease for which antiplatelets are indicated), but our analysis could not identify such an explanation. This paradoxical finding certainly requires confirmation and further study.

Frailty is included as a ‘yellow’ parameters in the EHRA 2018 Practical Guide, and hence, in combination with other yellow factors, can justify an appropriately reduced dose according to this system.

Regarding factors related to NOAC overdosing, our study identified lower renal function and AF patients who were not treated with apixaban. This can be explained by the fact that prescription of a reduced dose of apixaban depends on the presence of a minimum of two out of three criteria (see Supplementary Table 1), which decrease the probability for an overdose. Renal function is a well-known risk factor as all NOACs are renally excreted and three of the four NOACs have absolute SmPC dosing reduction criteria depending on renal function [22].

Noteworthy, when reviewing the patients taking a NOAC concomitant with antiepileptic drugs (which can lower NOAC plasma concentrations), three patients (75.0%) were inappropriately dosed as classified by the two systems and one patient was appropriately dosed according to the SmPC but potentially underdosed according to the EHRA 2018 guide (apixaban 5 mg plus valproic acid, ‘dark brown’). This reflects the unawareness of the interaction of anti-epileptics with NOACs among clinicians, and the almost full absence of data on the clinical effect of plasma-lowering medication on the efficacy of NOACs. Further Phase 1 studies are needed in which NOAC plasma concentrations may be better defined under these combinations.

6.3 Primary Care Versus Specialist Care

Although one of the initial objectives of this study was to investigate the prescription patterns in primary care versus cardiologist care, well-founded conclusions cannot be made due to the underpowerment as the result of the cessation of inclusions by the COVID-19 pandemic. Moreover, data would need interpretation in the light of the different patient demographics, such as age and renal function (Table 1). These two parameters are critical factors in the dosing criteria of both SmPC and EHRA 2018 guides. Nevertheless, there is a higher rate of inappropriate dosing in GP care compared to cardiologist care, which seems mainly the result of inappropriate overdosing. Overall, inappropriate NOAC dosing in primary care in Portugal, Belgium and the UK has been reported by other investigators in a range between 18.3–30.3% [25, 28, 29]. So far, a proven difference with specialist care is lacking from the literature, although such findings might be important to tailor and focus educational initiatives.

6.4 Influence of Different Renal Function Formulae

Although the Cockcroft–Gault renal formula was used in all the landmark NOAC trials, and hence adopted in the SmPC guidelines and EHRA guide, laboratories cannot routinely report this value since they miss out information such as patient weight, and rather report estimated glomerular filtration rate (eGFR) based on the MDRD or CKD-EPI formulae. A post hoc analysis using these two eGFR formulae showed no significant impact on the classification of appropriateness according to the SmPC and EHRA 2018 Practical Guide, although a slightly higher proportion of patients received a non-significantly inappropriate NOAC dose when MDRD or CKD-EPI were used. Hence, recalculating renal function using the CG formula, especially in AF patients with borderline eGFR, could be helpful to improve prescription correctness among clinicians. Other studies in larger AF cohorts also investigated the influence of eGFR formulae on dosing appropriateness and recommended using the CG formula in patients with a GFR < 70 mL/min and/or elderly patients aged ≥ 75 years [30, 31].

6.5 Limitations

Several limitations have to be acknowledged. An important limitation was the underrepresentation of primary care patients, as already mentioned. Atrial fibrillation patients included at the cardiology outpatient clinic originated from one centre, which limits generalisability, although they were recruited from the different cardiology sub-specialty clinics. Some primary care patients could be in regular follow-up by other cardiologists than those of the Antwerp University Hospital. The size of our cohort did not allow for analyses of each NOAC separately. The same applies to the multivariate results, which need to be interpreted with caution. Furthermore, as this was a retrospective quantitative study, based on the factors for NOAC dose adaptation included in the SmPC and 2018 EHRA Practical Guide, other possible influencing factors could not be objectified. Additional prospective (qualitative) research in specialist and primary care can provide more insights into dosing decisions and improving AF care. Finally, the EHRA Practical Guide and its dose-adjustment chart has to be regarded as a guidance tool to support clinicians in rational decisions, although definitive evidence on outcomes is not yet available and further studies are needed.

7 Conclusion

Inappropriate NOAC dosing in AF patients in follow-up by cardiologists and primary care physicians still occurs regularly, i.e., in about one in five patients (19.4%), according to the SmPC. Based on the 2018 EHRA Practical Guide, this proportion is significantly lower (15.6%), likely because more complex patients can be accounted for, but it is still very high. This calls for further physician education, a structured and frequent reassessment of NOAC dosing in complex AF patients, and further investigation on what might be appropriate dosing in very specific patient situations.