Right drug, wrong dosage: insights from the PAVE-AF antithrombotic study in older patients with atrial fibrillation

The PAVE-AF (Pattern of Antithrombotic treatment in the Very Elderly with Atrial Fibrillation) study was a prospective, multicenter, observational study that recruited consecutive patients ≥80 years with NVAF from 30 participating centers in Greece (15 hospitals and 15 private practice centers). The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the respective institutional review boards. Outpatient hospital clinics (university or state, located in either major urban centers or small towns) and private practice cardiologists were included based on their geographical distribution and in an attempt to obtain a representative sample of treating physicians. In order to maintain a nationwide, representative sample, the number of consecutively enrolled patients in each geographical region in Greece was calculated as a percentage share of the total sample size. The respective percentages were defined based on the distribution of the resident population aged ≥ 80 years by geographical region, according to the latest revision of the 2011 Population - Housing Census in Greece as completed by Hellenic Statistical Authority (supplementary Fig. 3, supplementary Table S3) [5].

Inclusion criteria were: (a) male or female patients with confirmed NVAF (excluding moderate to severe mitral stenosis or mechanical prosthetic valves) and (b) age ≥ 80 years. All patients signed an informed consent before their enrolment. AF had to be documented either on surface ECG or on Holter recording with a minimum duration of 30 s. AF definition and classification were standardized based on the latest ESC guidelines [6].

Data were recorded in an electronic case report form (eCRF) and included socio-demographic information, comorbidities and type of treatment. Uncontrolled hypertension was documented in the presence of systolic blood pressure > 160 mmHg, while labile INR was defined as unstable INRs or time in therapeutic range < 60%. Furthermore, the most recent blood test of creatinine levels (within 3 months) was recorded and creatinine clearance (CrCl) was calculated based on the Cockroft-Gault equation. Patients’ actual body weight was included in the respective formula. If patients received either AP only or no treatment for stroke prevention, then physicians were asked to select the reason(s) that led to this decision.

Patients treated with NOACs were classified into three dosing categories (recommended, underdosing and overdosing). Based on the approved European labels, the recommended dosing regimens were the following: Apixaban: 5 mg twice daily (bid); dose reduction to 2.5 mg bid if CrCl 15–29 ml/min, or if at least two out of three criteria fulfilled: age ≥ 80 years, body weight ≤ 60 kg, creatinine ≥1.5 mg/dL; contraindicated if CrCl <15 ml/min. Rivaroxaban: 20 mg once daily (od); dose reduction to 15 mg od if CrCl 15–49 ml/min; contraindicated if CrCl <15 ml/min. Dabigatran 110 mg or 150 mg bid if CrCl >30 ml/min; 110 mg bid in patients ≥80 years or at high risk of bleeding based on the summary of product characteristics; contraindicated if CrCl <30 ml/min.

Since included patients were ≥ 80 years, all apixaban-treated patients met one criterion for dose reduction, while the recommended dose for all dabigatran-treated patients was 110 mg bid. Patients with a contraindication to the administered NOAC due to reduced creatinine clearance levels were classified in the “overdosing” category. All data were acquired during a single visit by the investigator of each center. No follow-up visits were performed.

We calculated that a sample size of 1020 patients would provide statistical power in the range of 88% to 98% with 5% type I error to accurately predict OAC prescription. Descriptive methods were used to characterize the distribution of continuous variables. For non-symmetrically distributed data, the median and interquartile range were calculated. Comparisons were performed using two sample t-test, ANOVA or Mann-Whitney test, as appropriate. For multiple comparisons a Kruskal-Wallis test was applied. Multivariate logistic regression analysis was conducted. In the case of the model selection procedures, the backward selection procedure was performed; an alpha value (level of significance) greater than 10% was used as exclusion criterion. Categorical variables were described by frequencies and related percentages per class level. Comparisons were performed using chi-square test.

The study population included 1018 patients [mean (SD) 85.4 (4.0) years, 54.4% females], with NVAF (paroxysmal 32.4%, persistent 8% and permanent 59.6%). In total, 899 (88.4%) patients received AC, 82(8%) were treated only with AP, while 37 (3.6%) patients received no treatment. Regarding AP treatment, 37 patients received aspirin, 32 clopidogrel, 2 triflusal, while 11 patients were treated with a combination of aspirin and clopidogrel. Patient characteristics in each treatment group are presented in Table 1.

The type of administered antithrombotic treatment differed significantly among AF type groups with a higher rate of AP treatment among paroxysmal AF patients (supplementary Fig. 4). Furthermore, the type of antithrombotic treatment differed significantly among age subgroups with a lower proportion of nonagenarians receiving anticoagulants (Table 2). Nonagenarians were less likely to receive an AC (OR 0.472) and significantly more likely to receive AP treatment compared to patients < 90 years (OR 2.333, p = 0.001). Similarly, more patients aged ≥9 0 years received no treatment versus patients < 90 years (data not shown).

In the multivariate analysis, the following factors were significantly associated with administration of AC treatment as compared to AP or no treatment: low (0–2) HAS-BLED score [OR 8.62, p < 0.001], hypertension [OR 4.85, p < 0.001], labile INR [OR 3.89, p = 0.003] permanent AF [OR 3.60, p < 0.001], absence of uncontrolled hypertension [OR 3.07, p < 0.001], prior stroke/systemic embolism [OR 2.78, p < 0.001], age (continuous variable – per year of age) [OR 0.87, p < 0.001] and male gender [OR 2.26, p = 0.001] (Fig. 1).

Physicians reported that the following reason(s) led to their decision to administer either AP only or no treatment for stroke prevention: anticipated high bleeding risk in 57 (31.2%) patients, patient denial in 56 (30.6%), inability of regular INR monitoring in 41 (22.4%), adverse event on prior anticoagulant treatment in 23(12.6%), inability to maintain INR within therapeutic range in 22 (12%), high treatment cost in 10 (5.4%) and other reasons in 20 (11%) patients.

Among 899 patients treated with anticoagulants, 21 received parenteral and 878 oral antithrombotic treatment. In the latter subgroup, 293 patients were treated with vitamin K antagonists (33.4%), while 585 (66.6%) received NOACs [304 (34.6%) apixaban, 83 (9.5%) dabigatran and 198 (22.6%) rivaroxaban]. A significantly lower percentage of patients ≥ 90 years were treated with NOACs as compared to patients aged 80–89 years (48.4% vs 59.2%, respectively, p < 0.05). The reduced NOAC doses were administered in 71.8% of apixaban-treated patients (2.5 mg bid), in 82% of dabigatran-treated patients (110 mg bid) and in 63.2% of rivaroxaban-treated patients (15 mg od). In total, 63.2% of patients received NOAC dosing consistent with European label recommendations, 29.7% received a lower dose, while 7.1% were overdosed (supplementary Table S4). The percentage of patients receiving recommended, over- or under- dosing differed significantly among the three NOAC groups (p < 0.001), (Fig. 2).

The latest ESC guidelines do not recommend antiplatelet monotherapy for stroke prevention in AF patients [6]. However, there has always been a considerable time lag before issued recommendations are incorporated into everyday practice. Several real-world surveys have reported the persistence of antiplatelet use for stroke prevention in AF patients [7‐10]. However, most registries do not provide detailed data regarding prescription patterns in older patients. Furthermore, reported findings pertain to a transient period following approval and initial market penetration of different NOACs. This dynamic state may have prevented the capture of a steady and representative prescription behavior of physicians for stroke prevention. Our study enrolled patients several years post launch of NOACs in the Greek market, where the existing pricing and reimbursement policy allows patient access to this pharmacological treatment, thus reducing potential confounding effect of drug availability and affordability. In our study we found that 8% of real-world AF patients over 80 years are treated only with antiplatelets for stroke prevention, while only 3.6% receive no treatment at all. The most frequently reported reasons why an anticoagulant was not administered included anticipated high bleeding risk, patient denial and inability of regular INR monitoring, while treatment cost influenced therapeutic choices in a minority of cases. Our study also demonstrated several independent predictors associated with administration of antiplatelet monotherapy or no treatment in older patients over 80 years with AF, including increased bleeding risk (high HAS-BLED score), older age, uncontrolled hypertension, female gender and history of paroxysmal AF.

NOACs have evidence-based credentials for stroke prevention in patients with NVAF [11]. Their advent has signaled a shift in the prescribing pattern of therapies for stroke prevention. A Danish nationwide registry showed a significant decrease in the usage of vitamin K antagonists from 92 to 24%, with a counterbalancing increase in NOAC use [12]. Real world data evaluating the use of NOACs in older AF patients are scarce. In our study, two thirds of patients treated with oral anticoagulants received NOACs, thus supporting increased utilization of NOACs over vitamin K antagonists.

Apart from the standard doses, reduced NOAC dosing schemes have been approved for predefined patient subgroups with certain baseline characteristics [11]. The administration of reduced NOAC dosing in patients not meeting the recommended criteria worsens patient outcome and significantly increases the risk of adverse events [13]. NOAC overdosing was reported to increase all-cause mortality, while underdosing increased cardiovascular hospitalizations as compared to recommended doses [13]. The reported incidence of inappropriate NOAC dosing varies. Steinberg et al. have reported that only 13% of AF patients treated with NOACs received off-label dosing [13]. This percentage is considerably lower than the respective in our study. We found that 29.7% of patients received a lower dose, while 7.1% were overdosed. Several factors may explain this discrepancy. The authors have commented that the reported results may underestimate the magnitude of incorrect dose selection in broader AF populations, due to the potential confounding effect of participation in an AF registry [13]. The ORBIT-AF II registry evaluated the US community practice and ascertainment of appropriate NOAC dosing was based on the FDA-approved package inserts that differ from the respective European labels. Thus, the use of different classification criteria limits the extrapolation of relevant findings in the European care setting. Furthermore, the ORBIT-AF II analysis included a general AF patient population, with a mean age of 70 years, while our study specifically focused in the older patient subpopulation over 80 years. The popularity of a “softer” anticoagulant prescribing practice among AF patients over 80 years due to concerns on frailty, bleeding and falling, may partly explain the increased rate of underdosing [14]. In line with our results, Barra et al., in a retrospective chart review of 224 patients with a mean age of 77.7 years who received a reduced NOAC dose, found that only 43.3% of patients satisfied the dose adjustment criteria recommended by the manufacturer labeling [15].

Our findings reflect physicians’ willingness to avoid bleeding complications in the context of a safety-driven decision-making in anticoagulant type and dosage. Interestingly, there is a shift in the adopted “soft” thromboprotective options from antiplatelet monotherapy to inappropriately low doses of NOACs. However, inappropriate NOAC dosing in the older AF patients still deprives them of optimal thromboprotection. Intensive educational campaigns are needed to further familiarize prescribing physicians with the indicated NOAC dosing schemes and to educate them on the unfavorable prognostic impact of inappropriate NOAC dosing.

Several methodological strengths of our study should be commented. PAVE-AF is a prospective non-interventional study which is considered to provide evidence of “higher quality” than retrospective claims analyses. The latter is a common source of real-world evidence, but prone to bias due to void selection of diagnosis coding, missing data and lenient case definition. In our study, AF was confirmed either by surface ECG or on Holter monitoring with a duration of at least 30 s. Furthermore, we used minimal exclusion criteria to increase the external validity of our results. The study population had a representative geographical distribution to reduce the likelihood of systematic enrolment bias and to increase the generalizability of our findings. The number of patients enrolled in each geographical region was proportional to the reference population (≥80 years), based on the latest national population census. The participating centers have been selected on the basis of including different care setting for AF patients including both physician offices, as well as different types of hospitals. In addition, consecutive sampling was employed with a predefined target size to enhance the representativeness of the reference population and all eCRFs were validated through source data verification processes. However, strict implementation of consecutive patient enrolment cannot be confirmed, since use of screening logs was not mandated by the study protocol; thus, selection bias cannot be excluded. Another limitation of the study is the lack of follow-up data that do not allow the evaluation of patient outcome.