During recent years, three new anticoagulants (dabigatran, rivaroxaban and apixaban) have been introduced to the market, probably with one more anticoagulant (edoxaban) in the next 2 years. This review is not intended to compare the efficacy and risks of these new agents, but rather to detail the advantages and limitations. The pharmacokinetic characteristics of these drugs have few drug and food interactions, predictable dose responses, and rapid onset and offset, thus resulting in simplified management of the patient requiring anticoagulant therapy. No routine laboratory monitoring is required. A somewhat unexpected, but exciting observation involving the new anticoagulants, is the uniform reduction in intracranial bleeding by one-half compared with warfarin. The potential limitations of the new anticoagulants include uncertainty regarding assessment of drug levels, safe drug levels for major surgery, management of major bleeding, renal dependence, multiple dose regimens, adherence in the absence of frequent monitoring and unknown, rare side effects that were not captured in the trials. This review should clarify some of these concerns.
In the field of clinical thromboembolism, we have experienced a giant leap in the improvement in antithrombotic therapy. Thirty years ago, unfractionated heparin was preferably administered as an intravenous infusion, and vitamin K antagonists (VKAs) were the only available agents. Of VKAs, warfarin had and still has the most widespread use. Both VKAs and heparin require frequent laboratory monitoring; heparin requires monitoring on a daily basis. The development of low-molecular weight heparin (LMWH) enabled unmonitored once-daily subcutaneous injections and outpatient management of many patients, particularly those with venous thromboembolism (VTE). Although LMWH replaced VKAs for the long-term treatment for VTE in patients with cancer, LMWH is still considered inconvenient due to the need for daily injections.
Unfractionated heparin and VKAs interact with multiple steps in the coagulation cascade, which makes the estimation of the effect complex. In addition, VKAs have a large number of drug–drug and food–drug interactions, and variability in metabolism and target effect, which are determined by genetic polymorphisms. This constitutes the background for a 40-fold difference in the interindividual maintenance dose requirement (0.5–20 mg per day) to achieve exactly the same effect, which is likely unparalleled by any other drug.
Despite these difficulties, physicians and specialized anticoagulation clinics have learned to manage treatment with warfarin, achieving impressive efficacy and acceptable safety. Warfarin provides a risk reduction for stroke prevention in atrial fibrillation (SPAF) of 62%, which is far better than the reduction of approximately 22% experienced with aspirin . For the prevention of recurrent VTE, the results are even more impressive, with approximately 90% risk reduction [2, 3].
The most feared bleeding on warfarin is intracranial haemorrhage, which occurs in approximately 4 per 1000 treated per year , and has a mortality of approximately 50% [5, 6]. This complication contributes strongly to the hesitation of many physicians to prescribe warfarin and of many patients to take the drug [7, 8]. Due to such concerns, nearly 50% of patients with an indication for anticoagulation for stroke prophylaxis in atrial fibrillation are not treated .
A new generation of anticoagulants
The first anticoagulant with high specificity came from nature. Hirudin, produced in the salivary gland of the leech, is an inhibitor of thrombin at the last enzymatic step of the coagulation cascade . The irreversible binding of hirudin to thrombin probably contributes to the increased risk of bleeding . Despite the availability of hirudin, the quest began for small molecules that would bind reversibly in the active site pocket of thrombin or factor Xa and that could be suitable for oral administration. Although the concept was clear, the initial molecular prototypes provided a suboptimal effect, such as efegatran with its slow, tight binding to thrombin . With the utilization of X-ray crystallography, the three-dimensional configuration of the active site pocket of thrombin and the complex formation with D-Phe-Pro-Arg-CH2Cl or other nonpeptide small molecules was revealed down to a resolution of 1.9 Å at the Max Planck Institute for Biochemistry . Thereafter, tailor-made molecules with a high association rate constant to warrant rapid inhibition of thrombin could be produced, starting with melagatran [developed by Hässle (subsequently Astra and Astra-Zeneca] in Sweden) ]. The drug became the first orally available thrombin inhibitor after a small modification to improve absorption (the prodrug was ximelagatran). Moreover, melagatran was shown in clinical trials to have similar efficacy and bleeding rate as LMWH in the prophylaxis against VTE after major orthopaedic surgery [15, 16] and was comparable to warfarin in stroke prophylaxis in atrial fibrillation (SPAF) [17, 18] and treatment for VTE [19, 20]. The advantage of melagatran was convenience in view of eliminating the need for parenteral administration or laboratory monitoring. Ximelagatran was withdrawn from all markets in 2006 due to liver toxicity, which subsequently turned out to be specific for this molecule and not a class effect. Nevertheless, we had the proof of concept and other coagulation factor-specific, orally available drugs were entering phase III trials. At the time of writing this review, the oral thrombin inhibitor, dabigatran, and the factor Xa inhibitors, rivaroxaban and apixaban, have been approved for several indications, and a third Xa inhibitor, edoxaban, will have the phase III trial results for SPAF and treatment for VTE available when this article is in print.
General principles for the new anticoagulants
The details of the biochemical, pharmacokinetic and pharmacodynamic characteristics of these drugs have been published elsewhere . Some important preclinical characteristics are shown in Table 1. The VKAs work indirectly via inhibition of vitamin K epoxide reductase and vitamin K reductase, with subsequent synthesis of dysfunctional coagulation factors with slow onset and offset, depending on the elimination and resynthesis, respectively, of functional coagulation factors . In contrast, the new anticoagulants exert their effect almost immediately and as fast as LMWH when injected subcutaneously. Thus, the need for initial overlap with heparin is not essential for these agents.
Table 1. Important preclinical characteristics of the new anticoagulants
The metabolism for the S- and R-enantiomers of warfarin, as indicated.
The half-life of the new anticoagulants is slightly longer than LMWH, but definitively shorter than warfarin (Table 1). Thus, with rapid offset and onset, there is usually no need for bridging anticoagulation with heparin in case of interruption for surgery, as often carried out with VKAs. The new anticoagulants are stopped 1–2 days before elective surgery, although with dabigatran, the interval is extended to 4 days in the case of moderate renal failure in combination with procedures with a high risk of bleeding (Fig. 1). The main exception when parenteral anticoagulation in peri-operative management is required is when oral medication is not possible, for example, due to bowel paralysis postoperatively.
Although some drug interactions have been identified, mainly due to inhibition of the efflux transporter, P-glycoprotein (P-gp), and for factor Xa inhibitor dependence on the microsomal enzyme, CYP3A4, most of the interacting drugs are not used extensively, as reviewed in Table 2. Thus, the effect of the new agents is much more predictable than VKAs, and routine laboratory monitoring has become obsolete. Caution, possibly with a reduction in the dose of some of the new anticoagulants, is recommended when the patient is on concomitant medication with a strong P-gp inhibitor. For a few drugs with strong inhibition of both P-gp and CYP3A4, treatment with factor Xa inhibitors is contraindicated (Table 3). For patients with difficulties achieving a stable effect from VKAs, or with variable international normalized ratio (INR) results, a switch to a new agent can provide great relief. Whereas this is advisable for patients with unstable INRs due to frequent treatment courses with drugs that interact with VKAs, for example antibiotics and prednisone, it can be problematic if the cause is poor adherence.
Table 2. Drug interactions with the new anticoagulants
Table 3. Concomitant medications constituting contraindications to the use of new anticoagulants
(Not approved outside of Japan yet)
HIV protease inhibitors
HIV protease inhibitors
The maintenance dose of warfarin varies 40-fold between the extreme patient variants (0.5–20 mg day−1) and even more if patients with warfarin resistance are included. For the new anticoagulants, the dose is uniform for the majority of patients. In the studies on prophylaxis against VTE, two doses of dabigatran were tested (150 and 220 mg daily) and the lower dose was preferable for patients who are elderly (>75 years) or have moderate renal dysfunction (estimated creatinine clearance 30–49 mL min−1) . Similarly, for the SPAF indication, all the approved new agents have recommendations regarding dose reduction in special populations based on post hoc subgroup analyses (dabigatran, 150–110 mg b.i.d.) or on predefined criteria, as applied in phase III trials (rivaroxaban, 20→15 mg daily; apixaban, 5→2.5 mg b.i.d.). For patients with initial treatment for VTE, there is no such dose reduction in certain subgroups. In contrast, there is a change in regimen for rivaroxaban after the first 3 weeks (15 mg b.i.d→20 mg daily) and for apixaban after the first week (10→5 mg b.i.d.) to counteract the anticipated pronounced hypercoagulability at the time of diagnosis of a new thrombus.
Lower risk of intracranial bleeding
Many patients with atrial fibrillation have not been prescribed anticoagulants for optimal stroke prevention, but take aspirin or nothing . As already mentioned, a common reason for this choice is the fear of bleeding  and the most devastating haemorrhage is intracranial, with a mortality of approximately 50% when associated with warfarin [5, 24]. This risk is real, amounting to 0.4% per year , but outweighed by the risk of ischaemic stroke without warfarin, typically for these patients at 4–5% per year . In the long-term treatment studies with the new anticoagulants, it became clear that all phase III trials demonstrated a reduced risk of intracranial bleeding compared with warfarin (Table 4) [26-28]. Although this difference was not statistically significant in the smaller and usually shorter duration VTE trials, the same trend exists [29, 30]. It is therefore possible to talk about a (positive) class effect of the new anticoagulants versus VKAs. A plausible explanation is that initiation of coagulation via tissue factor (abundant in brain tissue) and factor VIIa pathway in response to bleeding is more compromised with VKAs, which markedly reduce the factor VII activity as opposed to the specific Xa or thrombin inhibitors .
Table 4. Incidence of intracranial haemorrhage during treatment with new anticoagulants or vitamin K antagonists in phase III trials
The clinical implications of these results cannot be ignored and should be an important tool to get a larger proportion of patients with atrial fibrillation to accept anticoagulant therapy. Likewise, for patients with VTE and at increased risk of recurrent events, it will probably be more acceptable to provide anticoagulation of indefinite rather than limited duration.
Potential problems with the new anticoagulants
Assessment of drug level
Particularly in emergency situations, there is a need to understand the current anticoagulant effect exerted by the drug. Typical situations are major bleeding and emergency surgery. Over the decades, clinicians have become accustomed to the INR, which provides a good correlation with the risk of bleeding related to VKAs . For unfractionated heparin and LMWH, there is a belief that the activated partial thromboplastin time (APTT) and antifactor Xa are useful, but in reality, the correlation with the risk of clinical events is poor .
Dabigatran levels can be reliably assessed quantitatively with the dilute thrombin time, which is commercially available as Hemoclot® (Hyphen BioMed, Neuville-sur-Oise, France) . For the standard dose of 150 mg b.i.d. and the sample obtained before next dose, a result >65 s, corresponding to >200 ng mL−1, is associated with an increased risk of bleeding. The assay is currently not widely available, and a rough quantitative assessment can be obtained with the APTT . The dose response is linear up to the therapeutic range, but the curve then flattens out, making the APTT unreliable for estimation of overdose. An APTT > 80 s (at trough) is associated with an increased risk of bleeding; with a normal APTT, the drug level is probably so low that surgery can be safely performed, although this will require verification with clinical data. Prothrombin time has a very flat dose response and is not helpful at all, whereas the regular thrombin time is too sensitive to be clinically useful.
Rivaroxaban can be qualitatively assessed with the prothrombin time, but unfortunately the sensitivity varies widely between the thromboplastin reagents; STA Neoplastin Plus (Stago, Asnieres, France) has been recommended as the most accurate assay . It is important to note that the results should be based on the prothrombin time and not converted to INR, which is specific for VKAs. For a quantitative estimation, chromogenic antifactor Xa assays can be used, but should be standardized for rivaroxaban. The rivaroxaban product monograph provides 5–95th percentile ranges for the prothrombin times in study populations receiving the different dose regimens used for various indications. It should be pointed out that there is no clinical data to validate how well results outside those limits correlate with adverse events. The APTT should not be used due to the flat dose response.
Apixaban levels are not reliably estimated with the prothrombin time or APTT. A chromogenic antifactor X assay, Rotachrom® Anti-Xa (Stago), provides a linear response curve and is recommended in the apixaban product monograph together with 5–95th percentile ranges with samples taken at peak or trough for different dose regimens, again without validation for clinically relevant correlates.
Edoxaban demonstrated a linear dose response with the prothrombin time and APTT in healthy volunteers . In summary, there is currently no available test that will be useful to estimate drug levels for all the new anticoagulants. In an emergency situation, it is imperative to obtain the correct name of the anticoagulant drug, which the patient is taking to select the appropriate test.
Lack of available reversal agents
Shortly after dabigatran was marketed, some alarming letters to the editor and editorials were published on the difficulties in managing bleeding in emergency situations and the elderly [36-38]. This triggered a class action suit in the USA and probably contributed to a slower than expected market gain for the drug. Against the arguments of the authors of the above quoted publications and the legal proceedings speak the following facts:
A lower incidence of intracranial haemorrhage with dabigatran than warfarin. This has already been discussed above and in Table 3. Furthermore, the consequences of intracranial haemorrhages are not worse when occurring on dabigatran versus warfarin because the fatal events are fewer in number .
When the resources required for managing all the major bleeds that occurred on treatment in five phase III trials with dabigatran versus warfarin were analysed, there were more red cell transfusions in the former group and equally more plasma transfusions in the latter group . The proportion of patients hospitalized and the average length of stay were similar, but the intensive care unit stay was actually shorter in the dabigatran group.
In the same pooled analysis, the outcomes after a major bleed did not differ between the groups regarding neurological function (assessed with the modified Rankin Scale) after intracranial haemorrhage or regarding the proportion of patients discharged home for rehabilitation or to long-term care after any major haemorrhage. Furthermore, there was a trend to lower 30-day all-cause mortality after a major bleed in the dabigatran group (P =0.057).
In an analysis of the subset of patients in the RE-LY trial who underwent surgery, there was no increase in major bleeding in the dabigatran group compared with the warfarin group, whether for elective or emergency procedures .
Real-world data are now emerging on the incidence of bleeding requiring hospitalization. The US Food and Drug Administration performed a sentinel project of drug surveillance and identified more intracranial haemorrhages (2.1–3.0 times), and in contradiction to the clinical trial results, more gastrointestinal haemorrhages (1.6–2.2 times) amongst patients newly prescribed warfarin versus dabigatran . This might have been biased by selective prescriptions of dabigatran to patients at a lower risk of bleeding. A Danish national registry study showed similar results , and a military healthcare database supported the findings regarding intracranial haemorrhage, but the incidence of gastrointestinal haemorrhage was higher for dabigatran .
Although a prothrombin complex concentrate is widely available in many countries and reverses elevated INR due to warfarin in 10 min, it is rarely used. In the above-mentioned review of major bleeds in the five phase III dabigatran trials, only a small number of patients in the warfarin group received this treatment for major haemorrhage .
In a Canadian study, as well as a multinational study, there was no evidence that treatment with a prothrombin complex concentrate improved survival after warfarin-associated intracranial haemorrhage [5, 44].
These arguments suggest that the lack of an available reversal agent for the new anticoagulants is more of a perceived than a real problem. Nevertheless, antidotes are under development and currently in phase II clinical trials. A humanized Fab fragment of a monoclonal antibody against dabigatran appears to give rapid and complete inhibition of the anticoagulant effect . For the direct Xa inhibitors, a recombinant decoy factor Xa (PRT4445) with similar conformation, but without haemostatic activity, reverses the effect of all drugs in this class .
Drug elimination with renal dependence
The proportion of active drug eliminated by the renal pathway is highest for dabigatran (Table 1), and patients with a creatinine clearance <30 mL min−1 (severe renal failure) were excluded from the trials involving this drug. Clinicians may be hesitant to prescribe dabigatran to patients with a creatinine clearance of 30–49 mL min−1 (moderate renal failure). The basis for this hesitancy is the fear of drug accumulation and an increased risk of bleeding. Subgroup analyses from all three SPAF trials with the new anticoagulants confirmed this trend (Table 5). Importantly, this pattern also pertained to the warfarin groups in these trials, and there was an increased risk of bleeding with LMWH and unfractionated heparin in patients with severe renal failure [47, 48]. In contrast, the risk of thromboembolism and stroke also increases with decreasing renal function, and thereby the benefit of optimal stroke reduction. Thus, we should not withhold effective anticoagulation from these patients. Most bleeds are extracranial, but all ischaemic strokes are intracranial.
Table 5. Incidence of major bleeding in the phase III trials in patients with atrial fibrillation according to calculated creatinine clearance and as defined in the trials [26-28]
Categories of renal function and major bleeding rate
In the studies with rivaroxaban and apixaban, a dose reduction was mandated in the protocol for certain risk groups, pertaining to most patients with moderate renal impairment.
The regulatory authorities have recommended annual reassessment of renal function. Renal function should also be evaluated whenever there is a risk of renal compromise, as follows:
decompensation of congestive heart failure;
increased dose of diuretics;
The calculation of creatinine clearance should be performed using the Cockroft–Gault formula  because other formulae (MDRD4 and CKD-Epi) underestimate renal failure in elderly patients. If the Cockroft–Gault formula is used, more patients would be identified with severe renal failure when dabigatran and other drugs would be contraindicated .
Poor adherence without INR monitoring
Patients on warfarin are monitored every 2–3 weeks, and the dose is adjusted according to the INR. For those patients who repeatedly demonstrate subtherapeutic levels, it is customary to evaluate whether or not the cause is a new interacting drug, increased intake of dark green vegetables, diarrhoea or poor adherence. In the latter case, the patient is usually warned about the consequences of inadequate stroke prophylaxis. In trials with new anticoagulant agents, adherence is most likely enhanced by the select, motivated study population, frequent follow-up visits and associated pill counts. Thus, in the RE-COVER study on treatment for VTEs , the adherence in the dabigatran group was 98.0%. Concerns can be raised that with dabigatran and apixaban, which are taken twice daily, the risk of poor adherence is higher than for once-daily rivaroxaban and edoxaban. Conversely, because the half-lives of the latter two are equally short or even slightly shorter, it can be argued that a missed dose may have worse consequences. Published data on other typical twice-daily medications (metformin, glipizide and metoprolol) show that adequate adherence, defined as at least 80% of pills taken, is only 52–65% [51-53].
A study was recently published on adherence with prescribed dabigatran in which careful initial instructions were given to the patients, and a return visit was scheduled after 3 months and then annually without any additional routine contacts . Adherence was checked using two methods (patient interviews and calculation of the proportion of days medication was dispensed for in relation to the number of days in that calendar period, with appropriate adjustments). The proportion of patients with adequate adherence combining both methods was 88%. It will be important to study the changes in adherence over time and in different clinical settings; however, based on these results, it is clear that with a well-informed patient, the adherence to a nonmonitored new anticoagulant is not necessarily bad.
No uniform regimen
It is well-known that the therapeutic range of VKAs is an INR of 2.0–3.0 in most cases; the major exceptions to this association are mechanical mitral valves or any mechanical valve together with atrial fibrillation . The dose of the VKA is thus adjusted to maintain this range. For most of the new anticoagulants, there are different doses for different indications. Rivaroxaban and apixaban require a change in the dose during the treatment for VTE. The elderly and/or patients with renal impairment need reduced doses. There is a geographical difference for dabigatran (75 mg instead of 110 mg in the USA). Unfortunately, medication dosing errors have occurred and will continue to occur. Physicians who prescribe new anticoagulants infrequently should become familiar with one of the agents and avoid using alternatives. Without head-to-head comparison studies, it is not possible to state that one agent is better than another.
New or unknown side effects
An observation was made in the dabigatran trial programme that myocardial infarction may be more common than with patients taking warfarin . This was first noticed in the RE-LY trial , although after additional, careful review, the difference was shown to be statistically nonsignificant . In the RE-MEDY study  on extended treatment after VTE, there was a significantly higher event rate on dabigatran compared with warfarin, but in the parallel RE-SONATE study, there was no difference versus placebo . Conversely, according to available data from the Danish registry , as well as the Military Healthcare Database , there was a decrease in myocardial infarction with dabigatran versus warfarin. It is likely that the debate on this matter will continue.
There is clearly always a concern with any new drug that rare side effects will be noted when very large numbers of patients have been exposed. Postmarketing studies, registries and sentinel projects have not disclosed any such events to date. Dabigatran now has a track record of >2 million patient-years, and rivaroxaban will have similar data in the near future. Any unknown toxic effects must be extremely rare. Nevertheless, physicians should not prescribe the drugs to pregnant women, who have not been included in any study; moreover, the small molecules most likely cross the placenta.
The advantages and limitations of the new anticoagulants are summarized in Table 6. There is no new drug that is entirely beneficial and to the satisfaction of everyone. Such is the case with the new anticoagulants. Some disadvantages are perceived, such as the lack of an available reversal agent, and other disadvantages are real, such as the need for appropriate renal function surveillance. The greatest advantages of the new anticoagulants include the convenience factor, the lack of need for routine laboratory monitoring and frequent dose adjustments and the reduced risk of the relatively rare, but ominous intracranial haemorrhages. Physicians need to become familiar with these agents, or at least to one agent, and use it appropriately for the safety of the patients. Anticoagulant therapy is potentially dangerous. Without adequate patient adherence or avoidance of concomitant medication with a P-gp- or CYP3A4-inducer, a poor anticoagulant effect can result in a higher incidence of stroke or recurrent VTE that occasionally is fatal. Failure to identify drug accumulation due to severe renal failure or medication with a strong P-gp- or CYP3A4-inhibitor can result in major bleeding. A well-informed patient is an important step towards improved net clinical benefit.
Table 6. Overview of the advantages and limitations of new anticoagulants
Heparin bridging not required
not for surgery
not for initial venous thromboembolism treatment (not tested for dabigatran or edoxaban)
Shorter time from last dose until operable or to bleeding cessation
Small interindividual variation in plasma levels
Standardized dose regimens
Few drug interactions
More predictable dose – no monitoring or dose adjustment
No food interactions
No food restrictions
Lower risk of intracranial bleeding
More acceptable treatment option for patients
PT and activated partial thromboplastin time (APTT) unsuitable for quantitative assessment of drug level
APTT and PT may be only be used for qualitative assessment for dabigatran or rivaroxaban, respectively. More specialized coagulation tests required for apixaban or for quantitative drug levels for any new anticoagulant
Drug level for safe major surgery unknown
Haematologist or other expertise should be consulted before surgery
Different coagulation tests required for different new anticoagulants
Requirement for education and for good communication with laboratory
No specific reversal agent
Half-life much shorter than for warfarin. Mainly, a problem in case of very high drug levels (severe renal failure, suicide attempt)
Renal dependence for elimination
Most pronounced for dabigatran. Renal function assessment annually or with intercurrent disease is required
Adherence difficult to ascertain
Patients with poor adherence to vitamin K antagonists regimen are unsuitable. Feedback from pharmacy on refill pattern might be helpful
No uniform dose regimen for all indications
Requirement for education, preprinted orders, electronic alerts
Conflict of interest statement
The author has received honoraria from Boehringer Ingelheim and Bayer Healthcare for work in study committees.