Oral Anticoagulation, Aspirin, or No Therapy in Nonvalvular AF
Oral Anticoagulation, Aspirin, or No Therapy in Nonvalvular AF
In this analysis of a large cohort of patients with CHA2DS2-VASc scores of 0 to 1, we show that those defined as low-risk (i.e., CHA2DS2-VASc = 0 [male], = 1 [female]) have a truly low risk for stroke, bleeding, and death. In addition, untreated AF patients with 1 additional stroke risk factor (i.e., CHA2DS2-VASc = 1 [male], = 2 [female]) are at increased 1-year stroke risk by 3.01-fold, bleeding by 2.35-fold, and death by 3.12-fold. Furthermore, in the patients with 1 risk factor, there were reductions in stroke and death with warfarin versus no treatment and with warfarin versus aspirin, but there was no increase in bleeding with warfarin versus aspirin. Sample size and low rates among treated groups led to some confidence intervals crossing neutral, as with stroke.
These data provide support for the approach advocated in the stroke prevention guidelines from the ESC and NICE, whereby the first decision step is to identify truly low-risk patients (CHA2DS2-VASc score = 0 [male], = 1 [female]) who have such low event rates that no antithrombotic therapy is recommended. Indeed, such patients do not have a positive net clinical benefit (NCB) from treatment, suggesting that antithrombotic therapy confers no advantage and, given the negative NCB, a possible disadvantage is evident. Our "real-world" data are also consistent with a modeling analysis by Banerjee et al., in which NOACs had a positive NCB in AF patients with ≥1 stroke risk factor. In large "realworld" cohorts, aspirin did not decrease stroke risk and demonstrated no positive NCB when balancing stroke against serious bleeding.
Subsequent to the initial step of identifying lowrisk patients, effective stroke prevention (i.e., oral anticoagulation) is recommended for patients with ≥1 additional stroke risk factor (CHA2DS2-VASc scores ≥1 [male] and ≥2 [female]). Our data support this recommendation, given stroke event rates >1.5 per 100 person-years at 1 year favoring warfarin treatment. Of note, aspirin resulted in a counterintuitive reduction in intracranial hemorrhage rates, perhaps explained by patient selection. Oral anticoagulation is effective stroke prevention, whether delivered as a NOAC or good-quality anticoagulation international normalized ratio control with a VKA (time-intherapeutic range >70%), which offers the best efficacy and safety when VKAs are used. Indeed, given the availability of NOACs, the threshold for treatment may be a stroke rate of 0.9% per year.
It has previously been shown that stroke rates could be as high as 8.1% among patients with a single risk factor (using the older CHADS2 score [congestive heart failure, hypertension, age ≥75 years, diabetes mellitus [1 point for presence of each], and stroke/transient ischemic attack [2 points] of 1) when substratified according to CHA2DS2-VASc score. Even among those with a low-risk CHADS2 score of 0, stroke rates, when substratified according to CHA2DS2-VASc scores, can range between 0.8% and 3.2% per year if left untreated. Indeed, when balancing stroke reduction against the potential for serious bleeding, a positive NCB for anticoagulation in AF patients with 1 stroke risk factor has been shown. The NICE guidelines clearly recommend that "aspirin monotherapy should not be offered for stroke prevention in AF patients," whereas oral anticoagulation should be offered to those with a CHA2DS2-VASc score ≥2 and offered to men with a CHA2DS2-VASc score of 1. This recommendation, therefore, is in alignment with the 2012 ESC guidelines.
Although the 2014 American Heart Association/American College of Cardiology/Heart Rhythm Society guidelines recommend oral anticoagulation for a CHA2DS2-VASc score ≥2, the recommendations for a CHA2DS2-VASc score of 1 state "no therapy, aspirin or anticoagulation". Female patients with a CHA2DS2-VASc score of 1 (by virtue of their sex) are low risk (hence, no antithrombotic therapy is necessary), but male patients with a CHA2DS2-VASc score of 1 would be left without stroke prevention if "no therapy or aspirin" were offered and would be at risk of fatal and disabling strokes. In a recent study of untreated AF patients with CHA2DS2-VASc scores of 1, the annual incidence of stroke was 6.6%, with the greatest risk conferred by hypertension and age 65 to 74 years. Even among patients unsuited for warfarin treatment, in whom the common practice previously was to offer aspirin, the oral factor Xa inhibitor apixaban was superior to aspirin for stroke prevention, with a similar rate of major bleeding and intracranial hemorrhage. Of note, other guidelines, including the previous U.S. guidelines, recommended oral anticoagulation to patients with 1 stroke risk factor (on the basis of the older CHADS2 score [22]), and both CHA2DS2-VASc and CHADS2 scores have many stroke risk factors in common. In essence, some risk factors weighted at 1 in the CHA2DS2-VASc score have the same score value in the older CHADS2 score, in which even the older guidelines recommend oral anticoagulation. Otherwise, such patients (evenwith a single stroke risk factor)will be exposed to unnecessary risk of disabling or even fatal thromboembolic complications or death that could otherwise have been prevented. Patients with AF are desperate to avoid strokes, and a recent survey on patients' values and preferences showed that patients were prepared to accept with oral anticoagulation 4.4 episodes of major bleeding just to prevent 1 stroke. In randomized trials, oral anticoagulation significantly reduces stroke and all-cause mortality, compared with use of control/placebo, and similar trends were seen in the current study for patients with ≥1 additional stroke risk factor.
We acknowledge the limitations of our observational cohort design, as with similar "real-world" cohort data. In Danish registries, the positive predictive value of the AF diagnosis is high (97%) (25), although our hospitalized patients with AF may have been a higher risk group for stroke and bleeding, and the applicability to communitybased (and often asymptomatic and "uncomplicated") AF cohorts is less uncertain. Our event rates probably differ from other published cohorts, given the dependence on definitions used for the particular dataset and endpoints, as well as populations studied (e.g., community vs. hospitalized), participation in a health care plan (vs. none), and trial (vs. nontrial) cohorts, among others. In addition, the type of AF (paroxysmal, persistent, or permanent) is not captured in Danish registries. Nonetheless, this study included patients with AF diagnosed both in-hospital and at ambulatory visits.
Although patients with AF in the low-risk stratum have a low rate of hospitalization, this finding is neither static nor predictable, and all published data show that hospitalization risks with AF are real (and increasing). Of note, the risk score assignment was made at baseline and was unaffected or not updated thereafter. Thus, during the first year, the CHA2DS2-VASc score (and its impact on event rates) was accurately estimated, but as time passed (i.e., for the full follow-up period), it may have become somewhat less accurate.
Some event rates in the nonanticoagulated group may be influenced by use of aspirin bought without a prescription, although the proportion of over-the-counter use is small, and it would be unlikely that patients recommended aspirin treatment on a regular basis would buy it in the pharmacy without reimbursement. Given the (nonsignificant) stroke reduction of 19% compared with control/placebo in historical trials, aspirin probably has only a small effect on stroke. The national prescription registry does not provide information on indications for treatment; hence, the aspirin-treated group may especially include patients given aspirin for other indications and perhaps not as a permanent treatment. This potential misclassification could have influenced the estimated rates.
Unmeasured confounding is inevitable in observational studies. Even though the Danish registers provide almost full coverage of history of hospital diagnoses and prescriptions, relevant information from general practice that may potentially explain some effects is still unavailable. Our registry design also does not provide details regarding drug changes over time nor of the quality of anticoagulation control (e.g., time-in-therapeutic range). We recognize that treatment adherence/discontinuation is an important consideration for patient management, but a thorough analysis is beyond the scope of this paper, which addresses the question of whether CHA2DS2-VASc would be best managed with "nothing, aspirin, or warfarin" on the basis of "real-world" observational data.
Discussion
In this analysis of a large cohort of patients with CHA2DS2-VASc scores of 0 to 1, we show that those defined as low-risk (i.e., CHA2DS2-VASc = 0 [male], = 1 [female]) have a truly low risk for stroke, bleeding, and death. In addition, untreated AF patients with 1 additional stroke risk factor (i.e., CHA2DS2-VASc = 1 [male], = 2 [female]) are at increased 1-year stroke risk by 3.01-fold, bleeding by 2.35-fold, and death by 3.12-fold. Furthermore, in the patients with 1 risk factor, there were reductions in stroke and death with warfarin versus no treatment and with warfarin versus aspirin, but there was no increase in bleeding with warfarin versus aspirin. Sample size and low rates among treated groups led to some confidence intervals crossing neutral, as with stroke.
These data provide support for the approach advocated in the stroke prevention guidelines from the ESC and NICE, whereby the first decision step is to identify truly low-risk patients (CHA2DS2-VASc score = 0 [male], = 1 [female]) who have such low event rates that no antithrombotic therapy is recommended. Indeed, such patients do not have a positive net clinical benefit (NCB) from treatment, suggesting that antithrombotic therapy confers no advantage and, given the negative NCB, a possible disadvantage is evident. Our "real-world" data are also consistent with a modeling analysis by Banerjee et al., in which NOACs had a positive NCB in AF patients with ≥1 stroke risk factor. In large "realworld" cohorts, aspirin did not decrease stroke risk and demonstrated no positive NCB when balancing stroke against serious bleeding.
Subsequent to the initial step of identifying lowrisk patients, effective stroke prevention (i.e., oral anticoagulation) is recommended for patients with ≥1 additional stroke risk factor (CHA2DS2-VASc scores ≥1 [male] and ≥2 [female]). Our data support this recommendation, given stroke event rates >1.5 per 100 person-years at 1 year favoring warfarin treatment. Of note, aspirin resulted in a counterintuitive reduction in intracranial hemorrhage rates, perhaps explained by patient selection. Oral anticoagulation is effective stroke prevention, whether delivered as a NOAC or good-quality anticoagulation international normalized ratio control with a VKA (time-intherapeutic range >70%), which offers the best efficacy and safety when VKAs are used. Indeed, given the availability of NOACs, the threshold for treatment may be a stroke rate of 0.9% per year.
It has previously been shown that stroke rates could be as high as 8.1% among patients with a single risk factor (using the older CHADS2 score [congestive heart failure, hypertension, age ≥75 years, diabetes mellitus [1 point for presence of each], and stroke/transient ischemic attack [2 points] of 1) when substratified according to CHA2DS2-VASc score. Even among those with a low-risk CHADS2 score of 0, stroke rates, when substratified according to CHA2DS2-VASc scores, can range between 0.8% and 3.2% per year if left untreated. Indeed, when balancing stroke reduction against the potential for serious bleeding, a positive NCB for anticoagulation in AF patients with 1 stroke risk factor has been shown. The NICE guidelines clearly recommend that "aspirin monotherapy should not be offered for stroke prevention in AF patients," whereas oral anticoagulation should be offered to those with a CHA2DS2-VASc score ≥2 and offered to men with a CHA2DS2-VASc score of 1. This recommendation, therefore, is in alignment with the 2012 ESC guidelines.
Although the 2014 American Heart Association/American College of Cardiology/Heart Rhythm Society guidelines recommend oral anticoagulation for a CHA2DS2-VASc score ≥2, the recommendations for a CHA2DS2-VASc score of 1 state "no therapy, aspirin or anticoagulation". Female patients with a CHA2DS2-VASc score of 1 (by virtue of their sex) are low risk (hence, no antithrombotic therapy is necessary), but male patients with a CHA2DS2-VASc score of 1 would be left without stroke prevention if "no therapy or aspirin" were offered and would be at risk of fatal and disabling strokes. In a recent study of untreated AF patients with CHA2DS2-VASc scores of 1, the annual incidence of stroke was 6.6%, with the greatest risk conferred by hypertension and age 65 to 74 years. Even among patients unsuited for warfarin treatment, in whom the common practice previously was to offer aspirin, the oral factor Xa inhibitor apixaban was superior to aspirin for stroke prevention, with a similar rate of major bleeding and intracranial hemorrhage. Of note, other guidelines, including the previous U.S. guidelines, recommended oral anticoagulation to patients with 1 stroke risk factor (on the basis of the older CHADS2 score [22]), and both CHA2DS2-VASc and CHADS2 scores have many stroke risk factors in common. In essence, some risk factors weighted at 1 in the CHA2DS2-VASc score have the same score value in the older CHADS2 score, in which even the older guidelines recommend oral anticoagulation. Otherwise, such patients (evenwith a single stroke risk factor)will be exposed to unnecessary risk of disabling or even fatal thromboembolic complications or death that could otherwise have been prevented. Patients with AF are desperate to avoid strokes, and a recent survey on patients' values and preferences showed that patients were prepared to accept with oral anticoagulation 4.4 episodes of major bleeding just to prevent 1 stroke. In randomized trials, oral anticoagulation significantly reduces stroke and all-cause mortality, compared with use of control/placebo, and similar trends were seen in the current study for patients with ≥1 additional stroke risk factor.
Study Limitations
We acknowledge the limitations of our observational cohort design, as with similar "real-world" cohort data. In Danish registries, the positive predictive value of the AF diagnosis is high (97%) (25), although our hospitalized patients with AF may have been a higher risk group for stroke and bleeding, and the applicability to communitybased (and often asymptomatic and "uncomplicated") AF cohorts is less uncertain. Our event rates probably differ from other published cohorts, given the dependence on definitions used for the particular dataset and endpoints, as well as populations studied (e.g., community vs. hospitalized), participation in a health care plan (vs. none), and trial (vs. nontrial) cohorts, among others. In addition, the type of AF (paroxysmal, persistent, or permanent) is not captured in Danish registries. Nonetheless, this study included patients with AF diagnosed both in-hospital and at ambulatory visits.
Although patients with AF in the low-risk stratum have a low rate of hospitalization, this finding is neither static nor predictable, and all published data show that hospitalization risks with AF are real (and increasing). Of note, the risk score assignment was made at baseline and was unaffected or not updated thereafter. Thus, during the first year, the CHA2DS2-VASc score (and its impact on event rates) was accurately estimated, but as time passed (i.e., for the full follow-up period), it may have become somewhat less accurate.
Some event rates in the nonanticoagulated group may be influenced by use of aspirin bought without a prescription, although the proportion of over-the-counter use is small, and it would be unlikely that patients recommended aspirin treatment on a regular basis would buy it in the pharmacy without reimbursement. Given the (nonsignificant) stroke reduction of 19% compared with control/placebo in historical trials, aspirin probably has only a small effect on stroke. The national prescription registry does not provide information on indications for treatment; hence, the aspirin-treated group may especially include patients given aspirin for other indications and perhaps not as a permanent treatment. This potential misclassification could have influenced the estimated rates.
Unmeasured confounding is inevitable in observational studies. Even though the Danish registers provide almost full coverage of history of hospital diagnoses and prescriptions, relevant information from general practice that may potentially explain some effects is still unavailable. Our registry design also does not provide details regarding drug changes over time nor of the quality of anticoagulation control (e.g., time-in-therapeutic range). We recognize that treatment adherence/discontinuation is an important consideration for patient management, but a thorough analysis is beyond the scope of this paper, which addresses the question of whether CHA2DS2-VASc would be best managed with "nothing, aspirin, or warfarin" on the basis of "real-world" observational data.
Source...