The databases were searched using the relevant terms. In the Cochrane Library, 121 citations were identified with the keyword searches listed above. From these searches one systematic review was identified providing a comprehensive review of randomized controlled trials (RCTs) evaluating oral anticoagulants for preventing strokes in patients with atrial fibrillation. Aguilar reviewed five RCTs involving older adults. ⁹

In Pub Med, using search terms “warfarin” AND “atrial fibrillation/drug therapy, limits of “Aged 65+ years”, “clinical trial, meta-analysis, practice guideline, randomized controlled trial, review, multicenter study, English text from 1966, 122 studies were found. Articles were rejected if titles indicated drug studies or comparisons, economics of treatment, opinions, or commentaries. Seventeen articles were retrieved.

In Scopus, using search terms “atrial fibrillation and warfarin and complications or risks” and “aging or elderly” limited to English language, 356 articles were found. Articles were rejected unless they were specifically related to complications of warfarin therapy. Editorials and opinions were excluded. The Cochrane systematic review was considered the starting point of the literature review. ¹ All articles predating this review were eliminated. Ten publications met the study criteria.

Two studies included an inpatient sample of older adults. Perera identified 207 patients diagnosed with atrial fibrillation with a mean age of 82.7. ¹ The sample was divided into 3 arms, those treated with warfarin, other anticoagulant and no anticoagulant. Each arm consisted of a subgroup of frail and non-frail participants. Three and six months after enrollment, participants were contacted by phone to report any complications. During hospitalization, prescription of warfarin to frail patients decreased by 10.7% and prescription of warfarin to non-frail patients increased by 6.3%. Analysis of the sub-groups revealed that the incidence of major hemorrhage over 6 months was 20 % (23% in the frail group and 16.9% in the non-frail group) which is the highest reported for a study of elderly with atrial fibrillation. Risk of complications was highly associated with frailty in this study. The high rate of complications may be due to their acutely ill and post hospitalization status. Also, the reliability and validity of the outcome data may be affected by self-report and memory.

Gage et al., conducted a retrospective review using the National Registry of Atrial Fibrillation II dataset created from 23657 anonymous patient records gathered by Quality Improvement Organizations for the National Stroke Project. ¹⁰ Medical records from all Medicare beneficiaries who were hospitalized and coded for atrial fibrillation were selected randomly and stratified by state,. The samples consisted of 1245 patients (mean age 83) with high risk of falls documented by their physicians in the medical records, 18261 (mean age 79) who did not have a documented fall risk and a “trial-like” sample of 3236 (mean age 73) who were younger and healthier. Thirty-three percent of the fall risk patients received warfarin at discharge compared to 48.9% on non-fall risk patients and 54% of the trial-like patients. The primary endpoint was subsequent hospitalization for an intracranial hemorrhage, based on hospital coding. The rates of intracranial hemorrhage were 2.8% in patients at high risk for falls, 1.1% in non-fall risk patients and 0.53% in trial-like patients. The large, national sample size provides greater generalizability. A limitation of this study is the possibility of incomplete records. Incomplete or inaccurate physician documentation that determined patients’ fall risk may also affect reliability. Despite the rate of complications, the study authors support the use of anticoagulants in patients at high risk for falls that are at moderate to high risk of stroke but not in patients at high risk of fall but with low risk of stroke.

Burton et al. also conducted a retrospective review of 601 patients previously identified as having atrial fibrillation by their primary physicians. ¹¹ Patients were selected by “invitation” to have their data reviewed. These data included baseline characteristics, strokes, bleeding events and international normalized ratio (INR) results. The cohort was divided into 2 subgroups: <75 and >75 years of age. Average patient age was 77 years. Both patients groups were within target INR range 68% of the time while treated with warfarin. Annual rates of severe bleeding complications and all bleeding complications on warfarin were 2.6% and 9% respectively. The incidence of severe bleeding was higher in those > 75 (10.1%) than those < 75 years (8.2%). Lost or missing data and the fact that all patients in this study were previously on warfarin may affect reliability of the findings. Also, the study lacked of reports of minor bleeding complications which could have skewed the results.

Two studies were conducted to determine the incidence of complications in older people in the initiation phase of warfarin. Johnson et al. conducted a retrospective cohort study (n=228, mean age=81) in which patients were identified by computer-generated list of all patients hospitalized, aged 76 and with a discharge diagnosis code of atrial fibrillation and those dispensed warfarin by the hospital pharmacy during the admission. ¹² Questionnaires were sent to determine reasons for starting and stopping warfarin, and to determine any bleeding episodes or strokes. The primary author followed up with a phone call for completion. Additional information was obtained from family, hospital medical records and primary physicians. The mean length of warfarin treatment was 27.9 months. Fifty-six percent started warfarin during the hospital admission. Eighteen percent of the cohort experienced major hemorrhage during warfarin treatment. Thirty percent of the cohort died within the study period which is representative of the level of frailty of this cohort. The annual event rate of major hemorrhage was 10%. The annual rate of life-threatening and fatal hemorrhage was 4.5% and 0.9% respectively. The high hemorrhage rate could be multifactorial. All patients in the sample were hospitalized indicating a sicker cohort. Warfarin was initiated in 56% of the patients regardless of the presence of contraindications. These patients would typically be excluded from other studies.

Hyek et al. conducted a prospective cohort study to define the rate of major hemorrhage in patients in the initiation phase of warfarin and define the risk of bleeding in the early phase of therapy. ¹³ The sample of 472 (mean age 72) were outpatients referred to the anticoagulation clinic and divided into age groups of > 80 years and < 80 years of age. Patients were followed through the first year of warfarin therapy. Use of aspirin and non-steroidal anti-inflammatory drugs (NSAIDS) was recorded. During the study, participants were within desired international normalized ratio (INR) range of 2- 3 58% of the total time on warfarin. The rate of major hemorrhage was 7.2% and rate of intracranial hemorrhage was 2.5%. Patients >80 years of age experienced higher rates of major bleeding compared with younger patients (13.08 versus 4.75 per 100 person years, which is significant at P=0.010. Risk of hemorrhage was increased when INR was greater than 4.0 and within the first 90 days of therapy. The higher incidence of hemorrhage could be explained by the advanced age of the sample and the restriction to patients in the initiation phase of warfarin. The zero attrition rate and accurate documentation of patient use of aspirin and NSAIDS reduces confounding of the study. Generalizability to the long term care setting is limited as patients in the long term care setting were excluded from the study.

In six studies, anticoagulation levels were monitored in designated anticoagulation clinics which likely contributed to tighter anticoagulation control resulting in a lower number of bleeding events. Poli et al. conducted two prospective studies. ^{14, 15} A 2007 study involved 290 patients (mean age 82) who were referred to an anticoagulation clinic. The sample was divided into groups as follows: 75-79 years, 80-84 years and 85-96 years. Two hundred fifty-one patients were newly started on warfarin. A computerized program monitored the INR. Comorbidities were assessed with patient interview and hospital records. Follow-up appointments documented warfarin dose, INR, other illnesses, hospital admissions and bleeding thrombotic events. The INRs were maintained at the therapeutic range of 2-3. Time spent within the therapeutic range of INR was 69%. The rates of total major bleeding were 1.4, 2.6 and 3.6 in the 3 groups respectively. Rate for cerebral bleeding was 1.35 for 100 patient years. Although these differences were not statistically significant, risk of bleeding increased with older age. The median INR related to bleeding events was 2.5. A higher risk of bleeding events was found in patients with diabetes. Complications of diabetes or medication prescribed for diabetes could be a factor due to the potential interactions with warfarin that may potentiate bleeding risk. ⁴ The low statistical power and the single clinic setting limit the study’s generalizability.

Poli et al. conducted another prospective study consisting of 783 nonvalvular atrial fibrillation patients who were referred for oral anticoagulation management to the anticoagulation clinic. ¹⁵ The study was conducted from June 1998 to December 2007. The median age was 75 years and 180 patients were 80 years and older. All patients were treated with warfarin with goal INR as 2 to 3. Warfarin dose was adjusted by using a computer-assisted prescription system. Patient interview, echocardiography results and hospital records provided patients’ demographic and clinical data. Patients spent a median of 14%, 71% and 15% of time below, within, and above the desired therapeutic range with no statistical significance between patients age <80 years and > 80 years old. During the study, 83 patients died, 9 from hemorrhagic complications. Forty-five of these patients were > 80 years old. Ninety-four patients suffered from bleeding complications: 37 major and 57 minor. The INR related to bleeding events was <3 in 67.5% of patients, between 3 and 4 in 19% in patients and > 4 in 13.5% of patients. Notably, major bleeding occurred at a median time of 36 months. Patients who were >80 years old with a history of previous ischemic event were associated with bleeding risk. History of hypertension or previous hemorrhage was not found to be associated with increased risk of bleeding which was presumed to be due to good blood pressure control in this sample. The authors of these two studies concluded that bleeding risk associated with warfarin administration is acceptably low in a clinic-monitored setting when anticoagulation levels are closely monitored. A strength of this study is the lengthy duration since major bleeding time occurred at a median time of 36 months. Studies that involve a shorter follow-up period may not accurately reflect bleeding complications.

Tincani et al. conducted a prospective cohort study (N = 90, Mean age 91.7) to assess the bleeding rates of extremely elderly patients with atrial fibrillation and examine risk factors that might predict bleeding. ¹⁶ All patients were enrolled during an ambulatory visit and followed for one year by the anticoagulation clinic. Patients enrolled had been on warfarin therapy for an extended period at time of enrollment. During clinic visits, parameters were obtained including INR, warfarin dose, history of other illness, hospital admissions and self-reported hemorrhagic and thrombotic events. Additional information was obtained from family, primary physician, hospital records and the laboratory as needed. Time spent within the therapeutic INR range of 2.5 was 66%. Potentially interacting medications were mentioned in the study. Two patients had intracranial hemorrhage and one patient had a minor hemorrhagic stroke (INR was 4.8 on day of the event). Two patients had major extracranial hemorrhage. Sixty-five percent of patients reported at least one bleeding episode. The author concluded that despite the elderly frail sample, risk of life-threatening bleeding was low and warfarin can be used to safely treat extremely elderly patients. Notably, the fatal hemorrhage was associated with an extremely high INR. Limitations in this study are the small sample size. Also, these patients had already passed the high risk initiation phase of warfarin which may explain the low rates of bleeding.

Torn et al. had similarly low bleeding rates in this three year retrospective cohort study n=4202 and were subtyped as follows: <60 n=842, 60-7- n=1200, 71-80 n=2742 and >80 n=1114. ⁷ The sample was selected from the Leiden Anticoagulation Clinic in the Netherlands. Follow up appointments involved gathering information regarding adverse events, hospital admissions, concurrent illness and blood sample for INR was collected. Data were also collected from hospital admissions, discharge summaries and autopsies. Outcomes were judged by a panel of experts who were blinded for the INRs. Time spent within therapeutic range (INR 3.0) declined slightly as patient age increased. Major bleeding incidence rose gradually with increasing age from 1.5% in those <60 to 4.2% in those > 80 years old. The incidence rate of fatal hemorrhage was consistent (0.3) for all age groups except those < 60 who had the lowest rate (0.1). The author mentions that the anticoagulation clinic as a study eliminates selection bias. However, inherent selection bias does occur as the patients were prescreened by the referring physician to initiate warfarin therapy. Knowledge of the percentage of individuals on long term warfarin therapy prior to the start of the study would also provide some explanation of the lower bleeding risks in this study.