Vitamin K antagonist use: evidence of the difficulty of achieving and maintaining target INR range and subsequent consequences

The consequences of INR instability are multifaceted. INR instability was associated with clinical events, higher level of medication non-persistence and discontinuation, utilization of more healthcare resources, and therefore, higher costs. According to meta-analyses of AF or mixed populations assessing INR control and associated events [18, 36‐38], greater than half of all thromboembolic events occurred when patients have an INR < 2.0, while over 40 % of all hemorrhagic events occurred at an INR >3.0. In VTE patients, subtherapeutic INRs were found to be present during 58 % of recurrent VTEs [17].

An observational study by Nelson et al, [38] using the Veterans Health Administration (VHA) dataset, explored the relationship between out-of-range INRs and clinical outcomes in 34,346 patients with non-valvular AF (NVAF) who were newly initiated on warfarin therapy. When INR values were below range (<2), patients were much more likely to experience adverse thrombotic or embolic events (Fig. 2). Patients were at an increased risk of major bleeding with both subtherapeutic INR values (RR = 2.58 95%CI: 2.19–3.03) as well as supratherapeutic INR values, (RR = 1.55, 95%CI: 1.21–1.97). All event rates were qualitatively the highest when patients had an INR < 2. While most of these events were stroke associated with sub-therapeutic values, increased bleeding events were also observed. While only speculative (and we could not identify supportive literature) it is possible the increased bleeding associated with sub-therapeutic INRs is due a lag in time between the actual event and the true INR value. This emphasizes the need for close INR monitoring to prevent subtherapeutic warfarin dosing.

Further evidence showed the link between INR instability and clinical events. In meta-analyses that examine the relationship between TTR and the prediction of adverse events, a significant negative relationship has been observed [19]. In patients with AF, 1 thrombotic or major hemorrhagic event per 100 patient-years could be avoided by improving TTR by 7 % or 12 %, respectively [19]. Likewise in patients with VTE, for every 1 % increase in TTR, recurrent thromboembolic events may be reduced by 0.46 % per year and major hemorrhagic events reduced by 0.30 % [17]. Furthermore, in a nested case control analysis of the Atrial fibrillation Clopidogrel Trial with Irbesartan for prevention of Vascular Events (ACTIVE W) study, patients who experienced an ischemic stroke had a TTR 9.5 % lower than those without any ischemic event [39]. The TTR of patients with a major hemorrhage was 7.2 % lower when compared to those without an event, again, suggesting that TTR is a useful predictor for both hemorrhagic and thromboembolic events. Of note, ACTIVE W also found that patients spent a greater amount of time out of range in the 1–2 months preceding a major bleeding event or stroke which suggests even a temporary period out of range can lead to a bleeding event or stroke.

Inability to achieve high TTR in clinical practice is associated with non-persistence and medication discontinuation [40]. In an analysis of longitudinal anticoagulation management records from 15,276 US patients with NVAF, discontinuation of therapy occurred in less than 4 months among patients with unstable INR. Patients who achieved INR stabilization were 10 times more likely to remain on warfarin therapy beyond 1 year. In another observational study using the Symphony Health Solutions’ Patient Transactional Database, patients who were prescribed rivaroxaban had a lower risk of treatment nonpersistence [HR 0.66 (95%CI: 0.60–0.72)] compared to patients who were prescribed warfarin [41]. A similar analysis of the Truven Health Market Scan Research Databases showed comparable findings, that NVAF patients who received rivaroxaban were 46 % less likely to discontinue therapy compared to those receiving warfarin [42]. Continued protection by anticoagulation is particularly important for patients with NVAF, since the risk of stroke is expected to increase with age and additional comorbidities [43].

INR instability was associated with higher healthcare utilization and costs. In an observational study using the Premier Perspective Comparative Hospital Database, hospital length of stay was 5.27 days vs. 4.46 days, leading to significant differences in hospitalization costs ($13,255 vs. $11,993, P < 0.001) [44, 45]. In another comprehensive cost analysis of 23,588 patients with NVAF who were on warfarin for at least 30 days from the US Veteran’s Administration, investigators randomly selected an INR value from a patient and classified it as being below 2, 2–3, or above 3 and then evaluated total direct costs (i.e. inpatient, outpatient medical, and outpatient pharmacy costs) over the next 30 days. Mean direct costs over 30-days after exposure to an INR <2.0, between 2 and 3, and >3.0 were $5126, $2355 and $3419 (Fig. 3) [46]. These findings remained robust in a sensitivity analysis with a more stringent definition of the cohort. The substantial cost difference between in-range and out-of-range time is significant across a broad warfarin population with atrial fibrillation.

The literature suggests that INR instability has important clinical and financial consequences which underscore the need for greater vigilance on achieving INR stability or the use of a novel oral anticoagulant which provides more consistent pharmacologic effects.

Based on data from adjusted meta-regression or multivariate analyses of large datasets, INR stability is known to vary greatly based upon various study- and patient-level factors (Table 3) [15‐17, 21, 47, 48]. The use of anticoagulation clinics can positively impact higher TTR attainment but only ~1/3 of VKA patients have access to these advanced services [49]. Therefore, a broad understanding of factors predicting INR instability is beneficial for clinical practice.

Two of the most extensive studies were conducted by Apostolakis et al. (SAMe-TT2R2) [14] and Rose et al. (VARIA) [48] and provided insight into factors affecting anticoagulation control. Apostolakis and colleagues [14] used data from the 1061 patients in the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) trial to identify clinical factors associated with TTR. Based upon these results, the SAMe-TT2R2 score was derived (and eventually validated) whereby 1 or 2 points are assigned for important patient factors (Table 4). Scores ≥2 were found to be associated with decreased odds of achieving a TTR ≥65 % (previously described as the minimum target TTR) [14].

The Veterans AffaiRs study to Improve Anticoagulation (VARIA) [48] used data from over 124,000 veterans receiving warfarin for any indication (55 % AF, 35 % VTE, 10 % other) between 2006 and 2008; and evaluated the effect of various patient characteristics on TTR in those starting warfarin (first 6 months of therapy) and who were experienced (on therapy for >6 months). Like the SAMe-TT2R2 derivation/validation study, female gender, younger age, minority status and co-morbid physical conditions were also found to be associated with lower TTR in VARIA (in both the inception or experienced cohorts) but there were a large number of additional factors which were identified, such as alcohol abuse, number of hospitalizations, and various comorbidities, such as heart failure, diabetes, chronic kidney disease, and others. The VARIA investigators also created a clinical prediction tool but eliminated race because they did not wish to perpetuate disparities in care, eliminated poverty and distance to drive to receive care because they felt it was hard to assess, and eliminated other factors to simplify the model. Their model is available in a downloadable excel spreadsheet from Supplemental Appendix 3S at http://​onlinelibrary.​wiley.​com/​doi/​10.​1111/​j.​1538-7836.​2010.​03996.​x/​full. In addition to the factors in the SAMeTT2R2 score, this tool also assesses the indication for use, total number of chronic medications, substance abuse, mental illnesses, number of hospitalizations, and the general quality of TTR attainment in other patients within that healthcare setting. Even with all of these additional factors, the R-squared value only ranged from 3.2 to 6.8 % suggesting that the much of the variability in TTR is not explained by this model. Furthermore, while the study assessed TTR at therapy initiation and after chronic therapy, the authors stated that the prediction tool is not to be used as a means to assess long term control, and that clinical experience from past VKA therapy is the preferred method [48].

Two reasons why the clinical prediction tools are inadequate may be related to genetics and adherence to therapy. Patient genotype plays an important role in INR stability [12, 50‐52]. At least 30 genes contribute to the anticoagulant effects of VKAs, with one third of the variance in warfarin dosing related to mutations in genes leading to the synthesis of CYP2C9 and vitamin K epoxide reductase (VKORC1) [12, 50‐52]. Patients with CYP2C9*2 and CYP2C9*3 polymorphisms have decreased enzymatic activity, metabolize warfarin (and to a lesser extent acenocoumarol) more slowly, have a 1.4- to 3.6-fold increased risk of supratherapeutic INR, and often take longer to achieve stable dosing [53‐55]. VKORC1 polymorphisms can result in either a heightened (group A haplotype) or reduced effect (group B haplotype) of warfarin which alters the risk of thromboembolism and bleeding accordingly [12]. Based on this data, the Food and Drug Administration altered the package insert recommending clinicians consider genetic testing before initiating warfarin therapy [56]. However, the cost-effectiveness of this approach is questionable; with economic models suggesting genotyping of patients would cost more than $170,000 per AF patient quality-adjusted life-year (QALY) gained (far above the commonly accepted willingness-to-pay threshold of $50,000 per QALY) [57].

Medication adherence was highlighted as an important variable in VKA INR stability in the American College of Chest Physicians (ACCP) guidelines [12]. Identified predictors of VKA nonadherence include not being married, not having a vehicle for transportation, education levels beyond high school, currently employed, lower levels of mental health functioning, poor cognitive functioning, and greater drug regimen complexity [58, 59]. In addition, studies have identified patient dissatisfaction with care as a cause of medication non-adherence in patients with cardiovascular disease states [60]. This is noteworthy since patient satisfaction with warfarin therapy has been shown to be poor in recent studies of AF [61] and VTE patients [62]. In the above-mentioned studies, poor patient satisfaction in either AF or VTE patients (measure by the Anti-Clot Treatment Scale) was related to the burden and frustration of taking VKAs resulting from fear of bleeding/bruising, diet and alcohol interactions and the perceived hassle of INR monitoring. One of the frequently cited studies evaluating the association between VKA non-adherence and INR control is the International Normalized Ratio Adherence and Genetics (IN-RANGE) Study [63]. IN-RANGE was a prospective cohort study conducted at 3 US anticoagulation clinics and assessed warfarin adherence using a medication electronic monitoring system (MEMS). The study followed 136 patients taking warfarin for a variety of reasons (but predominantly AF and VTE) for a mean of 32-weeks and found that missed warfarin doses (missed MEMS bottle openings) were common, with 92 % of patients missed at least 1 dose, and one-third missed more than 20 % of their doses. A total of 1490 INRs values were collected, with 40 % out of range and 26 % being below range. Upon multivariable regression analysis, researchers found that for every 10 % increase in missed warfarin doses (days without a dose), there was a 14 % increase in the adjusted odds of under-anticoagulation (having an INR < 2.0); and patients who missed >20 % of their doses (missed 20 days of warfarin therapy) had a 2.10-fold (95%CI: 1.48–2.96) increase in their odds of having an INR < 2.0 [63].

Existing research provided good understanding of the drivers behind poor INR control. The research findings indicate that many of the patient factors are not modifiable and are also not sufficiently reliable to predict whether VKA will perform well for a particular patient.