Introduction

Atrial fibrillation (AF) is the most frequent cardiac arrhythmia, with greater prevalence in older persons and patients with comorbidity such as diabetes mellitus, hypertension, heart failure, chronic kidney disease, and valvular heart disease [1, 2]. AF is associated with an approximately 5-fold increase in the risk of stroke, accounting for every fifth ischemic stroke [1, 3, 4], which may result in a poor quality of life, morbidity, and mortality [5,6,7,8]. Furthermore, delayed initiation of therapy associated with substantial stroke risk was reported [9]. For many years, vitamin K antagonists (VKA) were the cornerstone for stroke prevention in patients with AF. However, their use is limited by the need for frequent monitoring, a narrow therapeutic range, and a restricted diet, which often results in reduced compliance and persistence with therapy [10,11,12]. Non-Vitamin K antagonist oral anticoagulants (NOACs) emerged as an alternative with comparable or superior efficacy and safety to VKAs for stroke prevention in patients with AF. Furthermore, NOACs have fixed dosing regimens, do not require routine monitoring or significant diet restrictions [13,14,15]. Thus, treatment guidelines prefer NOACs over VKAs in patients with AF; however, routine shifting of stable VKA-experienced patients is not recommended [16,17,18]. Although NOACs were rapidly adopted as initial therapy based on reports derived from registries and guidelines, they often remain underutilized and often delayed or among patients with AF [9, 16, 19,20,21,22,23,24,25]. Previous studies have shown a 10–70% shifting rate from VKA to a NOAC [19, 26, 27] associated with younger age, female gender, eGFR ≥ 60 ml/min, higher CHA2DS2-VASC score, and specific comorbidities (stroke, bleeding, heart failure, and alcohol abuse). However, data dealing with initiation and shifting patterns and timelines remain scarce. The current study aimed to investigate patterns, predictors, timelines, and temporal trends of shifting from VKAs to NOACs.

Materials and methods

Study design and population

The present retrospective observational analysis utilized real-world data of the Maccabi Healthcare Services (MHS), one of four major state-mandated healthcare providers in Israel, serving 2.3 million members.

The database search for inclusion criteria included all patients insured by MHS at the time of the dataset closure. Patients considered for inclusion in the study were those 21 years of age or older with a documented diagnosis of AF (International Classification of Diseases version 9 [ICD-9] code 427.3x) who met at least one of the following criteria: (a) AF documented in either a hospital discharge form or cardiologist visit (b) two documented AF -related visits by a primary care physician, or (c) AF documented by any other physician and a documented prescription for either a VKA or NOAC. Eligible patients not prescribed either VKA or NOAC were not included. Of these patients, those with new diagnoses of AF during the screening timeframe of January 1, 2012, to December 31, 2015, who had never received oral anticoagulation therapy were included in the study and labeled the “incident cases” group. Patients, who had received the diagnosis of AF and had already been treated with VKAs for at least six months before January 1, 2012, were also included in the study and labeled as the “prevalent cases” group. The date of AF diagnosis was defined as the index date. Exclusion criteria: patients with known congenital heart disease, valvular heart disease, active malignancy, pulmonary embolism within six months from the index date, or pregnancy ending within 16 months before AF diagnosis. Follow-up continued until July 2016.

The study protocol was approved by the Maccabi Healthcare Services Institutional Review board (0046-16-BBL) and conducted in accordance with the Helsinki declaration. Patient consent was waived by the review board.

Israel has a National Health Insurance Law which mandates medical insurance for all citizens. Penetration of NOAC to the market was thus dependent on government approval. Medications covered under the “health services basket” are determined yearly by Israel’s Ministry of Health, and copayment for these medications is minimal. NOACs were included in this basket since 2011, but patients prescribed NOACs who did not meet the basket’s criteria could purchase these drugs with higher copayment based on their insurance plan. The criteria and indications stated in the yearly “health services basket” for NOACs were broadened over the years. NOACs can be prescribed by any physician, but the copayment necessitates an approval of the healthcare provider.

Data collection and definitions

MHS databases were used to collect patient and physician demographics. Socioeconomic status (SES) was derived from these databases by utilizing a commercial index (Points Location Intelligence, Ramat Gan, Israel) which was previously shown to highly correlate with the SES index provided by Israel’s National Bureau of Statistics [28]. Baseline comorbidities and risk score calculations were based on automated patient registries maintained by MHS, such as the diabetes mellitus [29] and cardiovascular diseases registries [30]. These registries are updated daily and automatically utilizing strict algorithms. Prescribing physician data, including birth year, sex, country of medical education attainment, and specialty, were collected from MHS’s human resources database.

The shift in therapy was defined based on dispenses. In addition, the patient had to have > 90 days of follow-up after the first dispense in that year to be eligible for this analysis. The time between AF diagnosis and first dispense (initiation) and the time from dispense of one group to dispense of a different group (shift) were collected and analyzed.

Statistical analysis

Patient characteristics were described by means, standard deviations (SD), and frequency (%). Intergroup comparison utilized the Mann-Whitney and two-sample t-test to compare between non-normally and normally distributed continuous variables. The Chi-square test was used for categorical variables comparison. The Kolmogorov–Smirnov method was employed to test for normal distribution. The Extended Mantel–Haenszel test for linear trend was used to assess trends in NOAC prescription between 2012 and 2015. Time to treatment shift from VKAs to NOACs, both in the prevalent and the incident populations, was described with Kaplan–Meier survival plots. Multivariate logistic regression was utilized to explore independent predictors of shifting. All variables introduced in the model were chosen based on clinical relevancy (forced model). Statistical analyses were performed with IBM SPSS Statistics for Windows, Version 25 (IBM corp. Armonk, NY). The study sample size allowed for 80% power for detecting variables associated with a shift to NOACs with a minimal adjusted odds ratio of 1.15 and a significance level of 0.05.

Results

Study cohort and initial prescription patterns

A total of 6987 eligible incident patients were identified throughout the study period. Of them, 2338 (33.4%) initiated treatment with a VKA and 2221 (31.7%) with a NOAC. In addition, 5259 prevalent patients were eligible for the current study. Baseline characteristics of the included patients and comparisons between patients first initiated with VKA vs. NOAC are presented in Table 1. Patients who were initially prescribed NOACs were older, more likely to be females, had higher socioeconomic status and CHA2DS2-VASc score and were less likely to be immigrants. Additional file 1: Figure S1 displays the temporal trends in the relative rates of the initially prescribed medication (NOAC vs. VKA) among the incident cases between 2012 and 2015. A statistically significant increase in the NOAC prescription rate was observed, from 32% to 2012 to 68.4% in 2015, with a corresponding decrease in the VKAs rate (p for trend <  0.001).

Table 1 Baseline characteristics of the study cohort; comparison between incident cases first prescribed VKAs to those first prescribed NOACs
Full size table

As presented in Fig. 2, the median time from diagnosis to first dispensing was greater in NOACs than VKAs; however, it decreased among patients treated with NOACs throughout the investigated period.

Shifting therapy

During follow-up, 3737 (49%) patients (54.3% and 47.1% of the incident and prevalent cases, respectively), initially prescribed with VKAs, shifted to NOACs. The clinical characteristics of patients, who shifted from VKAs to NOACs compared with those who did not, are presented in Table 2. Patients who shifted were older, more likely to be females, had a greater prevalence of congestive heart failure, diabetes mellitus, hypertension, hyperlipidemia, higher CHADS2 and CHA2DS2-VASC score, and were more frequently incident cases versus those that did not shift. The median time from initiation of therapy to shifting was 22.1 months [IQR, 8.3–37.6] and 39.3 months [IQR, 20.7–55.8] for the incident and prevalent cases, respectively and decreased over time during the investigated period (Fig. 1). The results of the multivariable adjustment for predictors of the shift from VKA to NOACs are presented in Table 3. Independent predictors included female gender, southern district, higher CHADS2, and CHA2DS2-VASC scores, younger age, non-smoking, and treatment with antiplatelets upon diagnosis. Figure 1 presents the rates of shifting from NOACs to VKAs among the two study groups. These rates have decreased over time from 8% to 4.5% in 2012 to 0.5% and 0.7% in 2015 in the incident and prevalent groups, p< 0.001, respectively (Fig. 2).

Table 2 Comparison of baseline and clinical characteristics between patients who shifted versus those who did not switch to NOACs during follow-up
Full size table
Fig. 1
figure 1

Time from initiation of treatment to shift (months, mean [IQR]) from VKA to NOACs

Full size image
Table 3 Mutually adjusted factors associated with shift to NOACS
Full size table
Fig. 2
figure 2

Time from initiation of treatment to shift (months, mean [IQR]) from VKA to NOACs

Full size image

Discussion

The current study evaluated the initiation and shifting patterns and timelines of NOACs vs. VKAs from a large healthcare provider in Israel. The main findings include the following: (1) A consistent increase in the rate of NOAC prescription and decrease in VKA prescription between 2012 and 2015. (2) The period from diagnosis to first dispensing was longer in NOACs than VKAs and decreased among patients treated with NOACs during the study period. (3) Shifting from a VKA to a NOAC was common (49%), occurring after a shorter time among the incident than prevalent cases, respectively, and decreased (the interval) throughout the study period. (4) Main predictors of such a shift were female gender, younger age, CHADS2 and CHA2DS2-VASC score, non-smoking, and antiplatelet therapy. (5) Shifting from a NOAC to a VKA occurred more frequently in incident cases and decreased over time.

The results of this study were stratified by two subgroups of AF (incident and prevalent cases). The baseline characteristics of these populations were different with incident cases, being younger and with lower risk for stroke as reflected by their CHA2DS2-VASc score. This finding can be explained by the increasing interest and awareness for stroke prevention seen with time and the emergence of NOACs as alternatives for VKA. The observed rates of initiation of NOACs throughout the entire study, as well as the significant increase in the proportion of eligible patients prescribed with NOACs, are consistent with previous literature (for an equivalent period) and are in line with the current guideline recommendation [1, 2, 16, 20,21,22,23, 27, 31,32,33,34,35,36,37]. However, consistent with previous studies, a significant proportion of patients are still not being prescribed an OAC for stroke prevention [27, 38,39,40].

The time from diagnosis to initiation of VKA therapy found in the current study was shorter than for NOACs and similar to the period reported by Khurshid et al. [9], where 86% of patients received a VKA. It is noteworthy that the stroke rate is substantial in the period between AF diagnosis and the administration of an oral anticoagulant (OAC) [9]. The delayed dispensing of NOACs in our study could perhaps stem from a more intricate approval process by the insurer, which has improved over the years, probably explaining the shortening of this gap. Other reported factors associated with delayed OAC dispensing include female gender, absence of hypertension, previous fall, and chronic kidney disease [9].

We found a significant shifting rate of about 50% from a VKA to a NOAC, more prominent in incident vs. prevalent cases. This rate is significantly higher compared to a study by Fosbol et al. [26] during a similar period (30%) and from those reported by Hale et al. [19] during 2010-2015. This discrepancy could result from differences in healthcare systems, particularly insurance and funding issues, such as the NOAC extended funding in our country. However, it is worth mentioning that although treatment guidelines recommend NOAC initiation over VKA in AF, routine shifts in stable VKA-experienced patients are not recommended [16]. The independent predictors of shifting from VKAs to NOACs are overall consistent with a previous report [26], mainly those comprising a higher CHADS2 and CHA2DS2-VASC score (except younger age) and an incident case rather than a prevalent case. These factors probably stem from a greater benefit of shifting with increased risk of stroke and from technical issues associated with insurance coverage (NOACs covered for higher CHADS2 score). Furthermore, when exploring etiologies for shifting from a VKA to NOAC, Hellfritzsch, et al., through a large cohort study, demonstrated that, at least in one out of five patients, shifting from VKA to NOAC therapy was preceded by a clinical event and subsequent need for medication re-adjustment [41]. Hale and colleagues reported that the main reason for shifting from VKA to NOAC treatment was NOACs being easier to use and manage [19]. The present study demonstrated that for patients on VKA shifting to NOAC, the time elapsed from initiation to shift was significantly longer in prevalent cases than newly-diagnosed patients and decreased in the latter group. This time lag may reflect the reluctance of physicians to switch an experienced and stable patient to NOACs.The shortening period in the incident (newly diagnosed) cases probably reflects the recommendation of NOACs as the drugs of choice in most patients with AF and the increased insurance coverage for this treatment in our country. Interestingly, a relatively novel finding was a non-negligible shifting from a NOAC to a VKA, which occurred more frequently in incident cases and decreased over time. The reasons are unknown but could result from contraindications developments for NOACs (mechanical valve implantation), adverse events, and financial limitations. The latter two, which can be more easily resolved with increased funding and types of NOACs, could explain the observed decrease trend in this shift over time.

The present study has several limitations. First, it is retrospective and observational, hence subject to inherent limitations of such a design. Specifically, patients’ preferences, which play a significant role in a real-world setting, could not have been evaluated. Second, diagnoses were primarily obtained from administrative databases (based on ICD-9), accordingly could be biased by coding errors. However, MHS continuously implements quality assurance measures. Third, the period for patient enrollment (2012–2015) might be considered limited for extensive, contemporary trends evaluation. Fourth, the entire exclusion of patients with a valvular disease might have excluded patients eligible for OACs. Fifth, OACs may have been prescribed to patients for indications other than AF (i.e., pulmonary embolism), and we lack the data to account for that. Yet, in the incident AF group, we included patients that were naïve to anticoagulation; in the prevalent AF group, we excluded patients with pulmonary embolism within six months from the AF. Thus, we believe that the potential bias is minimal. The fact that the included patients were diagnosed with AF and were eligible for OAC by their risk profile suggests that OAC prescription was aimed for stroke prevention in this population. Sixth, the different subtypes of NOACs and respective trends were not evaluated separately. Lastly, the Israel Ministry of Health co-sponsors medications purchase and co-payment of NOAC had changed over the years and was not identical to all patients due to different health plans, which could introduce bias. However, by including SES in our analyses, we believe it accounts for a partial adjustment for these differences.

Conclusions

A consistent trend of increase in the rate of NOAC dispensing and decrease in VKA dispensing was observed over time. The period from diagnosis to first anticoagulant dispensing was longer in NOACs compared with VKAs and decreased in both throughout the investigated period. Shifting from VKAs to NOACs occurred in 50% of the cases, more frequently among incident cases, with a decreasing time from therapy initiation to switching. The main predictors of such a shift were female gender, younger age CHADS2, and CHA2DS2-VASC score, non-smoking and concomitant antiplatelet therapy. Shifting from a NOAC to a VKA occurred to a lesser extent, more frequently in incident cases, and decreased over time.

Clinical implications

Upon diagnosis of AF or evaluation of a new patient with AF, prompt consideration of eligibility and indication and the patient-specific risk of OAC in general and NOAC specifically (as the class of choice in most patients) should be performed. When a decision to initiate therapy is made, it should be implemented with minimum delay. Caregivers and decision-makers should be aware of the reported shifting patterns when considering the initiation of NOACs. A socially and economically sensitive program to optimize the initiation and shifting of OAC therapy could significantly impact patient care and outcomes.