To Treat or Not to Treat
Anticoagulants as Secondary Preventives to the Oldest Old With Atrial Fibrillation
Background and Purpose—Anticoagulant treatment is effective for preventing recurrent ischemic strokes in patients who have atrial fibrillation. This benefit is paid by a small increase of hemorrhages. Anticoagulant-related hemorrhages seem to increase with age, but there are few studies showing whether the benefits of treatment persist in old age.
Methods—For this observational study, 4 different registers were used, among them Riksstroke, the Swedish Stroke Register. Patients who have had a recent ischemic stroke, were 80 to 100 years of age, and had atrial fibrillation, were included from 2006 through 2013. The patients were stratified into 3 age groups: 80 to 84, 85 to 89, and ≥90 years of age. Information on stroke severity, risk factors, drugs, and comorbidities was gathered from the registers. The patients were followed with respect to ischemic or hemorrhagic stroke, other hemorrhages, or death.
Results—Of all 23 356 patients with atrial fibrillation, 6361 (27%) used anticoagulants after an ischemic stroke. Anticoagulant treatment was associated with less recurrent ischemic stroke in all age groups. Hemorrhages increased most in the ≥90-year age group, but this did not offset the overall beneficial effect of the anticoagulant. Apart from age, no other cardiovascular risk factor or comorbidity was identified that influenced the risk of anticoagulant-associated hemorrhage. Drugs other than anticoagulants did not influence the incidence of major hemorrhage.
Conclusions—Given the patient characteristics in this study, there is room for more patients to be treated with anticoagulants, without hemorrhages to prevail. In nonagenarians, hemorrhages increased somewhat more, but this did not affect the overall outcome in this age stratum.
The benefit of anticoagulant treatment in patients with atrial fibrillation (AF) was shown in the 1990s. The role of anticoagulants in secondary prevention was confirmed by the EAFT (European AF Trial)1 and the Veterans Affairs Stroke Prevention in Nonrheumatic AF.2 With today’s aging population in the Western world, the mean age of the participants in these studies was relatively low, ≈70 years. This is important because the prevalence of AF increases with age. But as several studies have found that increasing age is a risk factor for major hemorrhage,3–5 many physicians are hesitant to use anticoagulants in high ages. The question is, therefore, to which extent the results of EAFT and Veterans Affairs Stroke Prevention in Nonrheumatic AF can be extrapolated to older patients.
A few studies have also included older patients. The BAFTA Study (Birmingham AF Treatment of the Aged)6 was a randomized controlled study that focused on patients >75 years of age (mean age, 81.5 years). In an observational study from Riksstroke, the Swedish Stroke Register, the mean age was 79.5 years.7 Both these studies demonstrate that elderly anticoagulant-treated patients have a more favorable outcome in comparison with patients who do not receive antithrombotic treatment or receive aspirin only.
With increased life expectancy and improved health, it is important to answer the question whether anticoagulant treatment can safely be used in even higher ages, without unacceptable increase of bleeding complications. Therefore, we analyzed observational data from national comprehensive registers, including 23 356 patients, 80 to 100 years of age, who had had an ischemic stroke and a history of AF. A second purpose for this study was to examine the association between the risk for hemorrhage and drugs commonly used in this patient group.
Patients and Methods
For this observational study, data from 4 Swedish national registers were linked, using unique personal identification numbers. First, the Swedish Stroke Register (Riksstroke), which includes data on >85% of all stroke patients in Sweden,8,9 identified the index stroke and provided data on baseline characteristics. Second, the National Patient Register (NPR), which includes data on diagnoses for all hospitalizations in Sweden, served as a complementary source for information on comorbid diseases. Third, the Prescribed Drug Register, which includes data on the purchase of prescribed drugs in Sweden, was used to ascertain that the medications were bought and picked up. The fourth source was the Cause of Death Register, which provided dates and causes of death.
From January 2006 through December 2013, 58 119 patients, 80 to 100 years of age, with an ischemic stroke (I63, International Classification of Diseases [ICD] Tenth Revision) were registered in Riksstroke and discharged alive. Of these, 23 356 (40%) had AF, diagnosed either before the stroke (according to the NPR) or newly diagnosed during the hospitalization (as reported in Riksstroke). The patients were stratified into 3 age groups: 80 to 84 (n=8850); 85 to 89 (n=9067); and ≥90 years of age (n=5439). All patients were followed until death or end of study, December 31, 2014. No patients were lost to follow-up with respect to date of death.
Riksstroke provided information on age, sex, and level of consciousness on admission (which served as a marker of stroke severity). Data on cardiovascular risk factors (ie, AF, hypertension [I10-15, ICD-10], diabetes mellitus [E10-14, ICD-10], and previous stroke [I61and I63-64, ICD-10]) were retrieved from Riksstroke and the NPR. Riksstroke was the source of information on smoking habits, and the NPR was the source of information on heart failure (I50; ICD-10). The NPR provided records on comorbid diseases (ie, ischemic heart disease [I20–25; ICD-10], heart valve disease [I05–08; I34–39 and Q22–23; ICD-10], cancer [C00–97; ICD-10], mental/behavioral disorder [F00–99; ICD-10], and previous major hemorrhage [I60–62, K25, K26, R31, S064, S065, S066, I850, K226, D629, K290, K920, K921, K922, H113, H313, H356, H431, M250, N421, N938, N939, and N950; ICD-10]). The assessment of major hemorrhage is previously presented in detail.10 Briefly, we defined major hemorrhage as the first event of bleeding necessitating hospitalization during the period of 1997 to the index stroke.
Information on the use of different medications (anticoagulants, antiplatelets, antihypertensives, statins, peroral antidiabetics, NSAIDs was retrieved from the Prescribed Drug Register. In Sweden, pharmacies expedite supplies of drugs for 90 days at a time. If a prescription was dispensed at the pharmacy within 120 days after the index stroke, we considered this medication to be in active use. Except for dabigatran (4.8%), warfarin was the only oral anticoagulant available at the time of the study. Antiplatelet agents included aspirin (96%), dipyridamole (6.8%), and clopidogrel (10%).
Outcomes were registered starting from 1 week after discharge until the end of follow-up. The 1-week time point was chosen to reduce the influence of early events associated with postponed anticoagulant treatment, to minimize discrepancy between the registers, and to reduce the influence of terminally ill patients who were discharged without anticoagulant treatment. Intracranial hemorrhage was defined as intracerebral, subdural, and subarachnoid hemorrhages, including traumatic hemorrhages (I60–62, S064-6). Major hemorrhage as an outcome was defined using the same ICD-10 codes as major hemorrhage as a risk factor.
All analyses, performed in agreement with privacy legislation in Sweden, were approved by the Regional Ethics Committee in Uppsala (registered number 2014/454).
The person-time at risk was calculated 1 week after discharge (index stroke; first registration in Riksstroke) until an event, death, or the end of study in December 31, 2014. The incidence rate per 100 person-years for subsequent events and death was calculated. Univariate and multivariate Cox proportional hazard regression models, with adjustment for single or multiple baseline covariates and the interaction between anticoagulant and antiplatelet agents, were used to calculate hazard ratios and 95% confidence intervals for subsequent events. The generalized linear model was used to obtain mean frequency of anticoagulant use adjusted for other covariates. The level of significance was 5%, and all the P values were 2 sided. All statistical analyses were performed using the Statistical Analysis Software, version 9.4, for Windows (Cary, NC).
During the years 2006 through 2013, 58 119 patients, 80 to 100 years of age, were registered with a first-event ischemic stroke in Riksstroke. Of these, 23 356 (40%) had AF. Overall, 61% of the patients with AF were women, and 39% were men, but the proportion of women increased with advancing age.
Of the 23 356 patients with AF, 6361 (27%) used anticoagulants at discharge from hospital. Most patients were on warfarin, but 305 (4.8%) were on dabigatran, which was the only novel oral anticoagulant available during the study period (2006–2013).
The baseline characteristics of all patients with AF, according to anticoagulant use at discharge from hospital, are presented in Table 1. The tendency to use anticoagulants decreased with increasing age and decreasing level of consciousness. There was also an unexplained sex difference. After adjustment for age, level of consciousness, and dementia, the prevalence of anticoagulant use was still 2.8% higher in men than in women.
In addition, there were differences between the anticoagulant users and nonusers in other characteristics, such as risk factors, comorbidity, and medication. Patients with dementia did less often use anticoagulants. Both antihypertensives and statins were more often used in the anticoagulant group.
Anticoagulants: Risk for Stroke and Death
Figure 1 shows hazard ratios for different outcomes (stroke, hemorrhage, and death) in relationship to anticoagulant treatment for the 3-age strata. The results are also available in Table I in the online-only Data Supplement. The incidence of ischemic strokes per 100 person-years was ≈40% to 50% lower in the anticoagulant group. Hemorrhagic stroke was almost 50% more common in the anticoagulant group, but as the number of hemorrhagic strokes was substantially lower than the number of ischemic strokes, the total incidence of stroke (ischemic and hemorrhagic) was still reduced by 40% to 50% in all 3-age strata. Figure 2 shows Kaplan–Meier curves on ischemic stroke for the 3-age strata according to anticoagulant use.
In all 3-age strata, there was a significantly reduced mortality in the anticoagulant group, both because of stroke and from all causes. Kaplan–Meier curves that show all causes of death by anticoagulant treatment for each age stratum, are given in Figure I in the online-only Data Supplement.
Risk for Hemorrhage With Anticoagulants
Hazard ratios for intracranial hemorrhage in patients treated with anticoagulants, by different age strata, are presented in Table 2. Although the effect does not reach significant levels, there is a trend toward higher risk for intracranial hemorrhage with advancing age in anticoagulant-treated patients. The effect of risk factors and comorbidities seem to be limited. Figure 3 shows the corresponding Kaplan–Meier curves in the 3-age strata. The risk for major hemorrhage (Figure II in the online-only Data Supplement) showed the same age trend.
Risk for Major Hemorrhage With Other Medications
In a separate analysis, we investigated the relationship between medications other than anticoagulants and major hemorrhage. We also analyzed interactions between antithrombotic drugs (anticoagulants, antiplatelets, and nonsteroid anti-inflammatory drugs) with respect to hemorrhage. There was no convincing relationship between any medication and major hemorrhage in any of the 3 age strata (Table II in the online-only Data Supplement), neither did we find any interaction between different antithrombotic drugs with respect to major hemorrhage (Table III in the online-only Data Supplement).
This observational study on patients ≥80 years of age with AF and ischemic stroke affirms the advantages of anticoagulant treatment also in this age group. Recurrent ischemic stroke was reduced in those who were on anticoagulants. Although hemorrhagic complications increased with advancing age in the anticoagulant group, the benefits of anticoagulants were not offset by an increased incidence of hemorrhages, not even in the highest age group (≥90 years).
Our study reports a few baseline differences between the anticoagulant and nonanticoagulant groups. The most important differences were age (which is the focus of this study) and the level of consciousness (a marker of stroke severity).11 Patients with more severe stroke were less often on anticoagulants, as was the case with patients having dementia. Although dementia was a relatively uncommon diagnosis, milder degrees of cognitive impairment may also have played a role in the choice of preventive medication. Together, these 2 factors, stroke severity and dementia, seemed to be the most decisive as to whether patients used anticoagulants. This may be because doctors believe that hemorrhages occur more frequently after a severe stroke and in patients who may not be able to follow a complicated regimen. Some may also argue that the patient’s quality of life is already poor, or they may feel that amyloid angiopathy associated with dementia may make the risk of hemorrhage much higher. In our experience, stroke physicians also fear that hemorrhages may occur more frequently with increasing multimorbidity. Taken together, these factors, and personal preferences, may determine different prescription patterns of anticoagulants.12,13
Ischemic strokes seem to be prevented even in the oldest old with the use of anticoagulants. Even if there was an increased risk of major hemorrhages in the highest age group, the increase of hemorrhages did not offset the benefits on ischemic stroke in any of the age strata. Our results are in agreement with the BAFTA study,6 although the primary outcome in their ≥85-year age group in that study was not significant. The BAFTA study found no increase of hemorrhages with anticoagulant treatment, although the number of events was few in the highest age groups. Other studies on anticoagulants in elderly patients found no critical increase of hemorrhages, provided that the prothrombin time levels were carefully maintained.14,15
We were not able to confirm the influence of cardiovascular risk factors or comorbidities on the risk of hemorrhage, as was found in some studies.16,17 Being a registry-based study, information on comorbidities may have been insufficient. Also, patients who were prone to bleed may not have been ordinated anticoagulant treatment from the start. With these reservations, the results suggest that the significance of comorbidities as promotors of hemorrhage should not be exaggerated because the protective effect of warfarin seems to prevail, at least in patients with no obvious contraindications to anticoagulant treatment.
The risk of major hemorrhage with drugs other than anticoagulants was also low in this material. Again, because this was an observational study, we must interpret this finding with some caution. Patients who have had supposed drug-related hemorrhage (eg, aspirin-related) were probably not ordinated that drug again. Our results can, therefore, be assumed to apply only to patients who have not had a drug-related hemorrhage.
This study has several strengths. It is a large study, and it uses the power of 4 national registers that have a high degree of coverage. The concomitant use of these registers is possible because of the presence of unique personal identification numbers in Sweden. The Drug Prescription Register includes data on purchase of prescribed products, which strongly indicates that a patient really uses that drug. The NPR has a decent positive predictive value and sensitivity for many of the comorbidities included in this study.18
Several weaknesses must also be acknowledged. First, this was an observational study. The anticoagulant and the nonanticoagulant groups were biased from the start. They differed in several ways, for example, stroke severity, dementia, and statin use. Although we have adjusted for the most important cofactors, this may not have been fully achieved. Although the level of consciousness mirrors stroke severity, the usage of a dedicated stroke severity scale, such as the National Institutes of Health stroke scale, would have been preferable. The proportion of patients with a valid National Institutes of Health stroke scale score has increased during the years but was still not >52% in 2013. This study involves secondary prevention using almost only warfarin. Separate studies need to be performed for anticoagulants in primary prevention and for the novel anticoagulants.
This study, which targeted stroke survivors between 80 and 100 years of age having AF, suggests that anticoagulants reduce the incidence of recurrent stroke in the oldest old. There were more ischemic strokes in those who were not treated with anticoagulants. In anticoagulant-treated patients, the benefit of treatment in any of the 3 age strata was not offset by hemorrhagic complications. Additionally, mortality was lower in the anticoagulant-treated group. Therefore, with the usage of anticoagulants that was prevailing in Sweden during the years in question, it is safe to say that anticoagulants were underused in this age group and that there is room for more patients on treatment.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.016902/-/DC1.
- Received February 1, 2017.
- Revision received March 23, 2017.
- Accepted April 4, 2017.
- © 2017 American Heart Association, Inc.
- Ezekowitz MD,
- Bridgers SL,
- James KE,
- Carliner NH,
- Colling CL,
- Gornick CC,
- et al
- Hughes M,
- Lip GY
- Hylek EM,
- Evans-Molina C,
- Shea C,
- Henault LE,
- Regan S
- Mant J,
- Hobbs FD,
- Fletcher K,
- Roalfe A,
- Fitzmaurice D,
- Lip GY,
- et al
- 9.↵Riks-Stroke. Riks-Stroke home page. http://www.riksstroke.org/forms/. Published 2015. Accessed February 1, 2016.
- Cohen N,
- Almoznino-Sarafian D,
- Alon I,
- Gorelik O,
- Koopfer M,
- Chachashvily S,
- et al
- Poli D,
- Antonucci E,
- Grifoni E,
- Abbate R,
- Gensini GF,
- Prisco D
- Rikala M,
- Kastarinen H,
- Tiittanen P,
- Huupponen R,
- Korhonen MJ