Bleeding and Recurrent Thrombosis in Definite Antiphospholipid Syndrome: Analysis of a Series of 66 Patients Treated With Oral Anticoagulation to a Target International Normalized Ratio of 3.5 | Clinical Pharmacy and Pharmacology | JAMA Internal Medicine | JAMA Network
[Skip to Navigation]
Sign In
Cumulative number of visits with the international normalized ratio (INR) within predefined ranges.

Cumulative number of visits with the international normalized ratio (INR) within predefined ranges.

Table 1 
Demographic Characteristics and Previous Clinical Manifestations (N = 66)*
Demographic Characteristics and Previous Clinical Manifestations (N = 66)*
Table 2 
Antiphospholipid Antibodies (N = 66)*
Antiphospholipid Antibodies (N = 66)*
Table 3 
Bleeding Rates
Bleeding Rates
Table 4 
Bleeding and Thrombotic Complications*
Bleeding and Thrombotic Complications*
Table 5 
Summary of Studies Including Cohorts of Warfarin-Treated Patients*
Summary of Studies Including Cohorts of Warfarin-Treated Patients*
Original Investigation
May 27, 2002

Bleeding and Recurrent Thrombosis in Definite Antiphospholipid Syndrome: Analysis of a Series of 66 Patients Treated With Oral Anticoagulation to a Target International Normalized Ratio of 3.5

Author Affiliations

From the Lupus Research Unit, St Thomas' Hospital, London, England (Drs Ruiz-Irastorza, Khamashta, Hunt, Escudero, Cuadrado, and Hughes), and Service of Internal Medicine, Hospital de Cruces, Bizkaia, the Basque Country, Spain (Dr Ruiz-Irastorza).

Arch Intern Med. 2002;162(10):1164-1169. doi:10.1001/archinte.162.10.1164

Background  Prolonged anticoagulation is the treatment of choice for patients with thrombosis and the antiphospholipid syndrome. However, there is still debate about the optimum intensity of anticoagulation.

Methods  The study included 66 patients with antiphospholipid syndrome (Sapporo criteria) and previous thrombosis. All were receiving oral anticoagulation to a target international normalized ratio of 3.5. Every patient was individually interviewed to recall major bleeding and thrombotic episodes during the previous 12 months.

Results  Patients were mainly women and white. The rate of major bleeding was 6 cases per 100 patient-years (95% confidence interval [CI] 1.6-15.0). The rate of intracranial bleed was 1.5 per 100 patient-years (95% CI, 0.04-8.4). None of the bleeding episodes was fatal. The rate of thrombotic recurrences was 9.1 cases per 100 patient-years (95% CI, 3.3-19.6). Most recurrences took place in the same vascular bed as the original thrombosis. Age, time receiving anticoagulant therapy, primary vs secondary antiphospholipid syndrome, positivity for anticardiolipin antibodies, positivity for lupus anticoagulant, previous arterial thrombosis, previous stroke, previous venous thrombosis, and previous thrombocytopenia were not predictive of bleeding events. However, the risk of thrombotic recurrences was independently higher in patients who were receiving anticoagulation for longer periods.

Conclusions  The risk of intracranial and fatal bleeding in patients with definite antiphospholipid syndrome and previous thrombosis treated with oral anticoagulation to a target international normalized ratio of 3.5 is similar than in groups of patients treated to lower target ratios. The risk of thrombotic recurrences, even during anticoagulation, was high. Most recurrences took place in the same territory as original thromboses.

SOON AFTER the description of Hughes (antiphospholipid) syndrome, it became clear that anticoagulant therapy was more effective than treatment with antiaggregant drugs for prevention of recurrent thrombosis in patients with antiphospholipid antibodies (aPLs).1-4 However, the optimum intensity of oral anticoagulation has since been a matter of debate. Two retrospective series including 701 and 1472 patients with the antiphospholipid syndrome (APS) and thrombosis showed a lower rate of recurrences when international normalized ratios (INRs) were higher than 3.0. In contrast, 2 small observational prospective studies of patients with histories of venous thromboembolism and aPL5,6 reported no thrombotic recurrences in patients at INRs between 2.0 and 3.0. As a result, definite conclusions are difficult to draw, and even recent guidelines for the treatment of patients with APS and previous thrombosis do not agree, recommending target INRs of 3.57 or 2.5.8

The heterogeneity of the different cohorts could explain in part these divergent results. For instance, patients included in the prospective studies5,6 had no history of arterial thromboses, a high prevalence of low-titer anticardiolipin antibodies (aCLs), and a lower risk of recurrences after warfarin sodium withdrawal compared with those of retrospective series.1,2 An additional major concern has been the increased risk of bleeding with high-intensity (INR, 3.0-4.0) oral anticoagulation. This study was an attempt to clarify the risks and benefits of oral anticoagulation to a target INR of 3.5 in patients with definite APS and previous thrombosis.

Patients and methods
Study design

This was a retrospective cohort study in which every patient included was individually interviewed by one of us (G.R.-I., A.E., or M.J.C.). The interview included specific questions to recall bleeding episodes and thrombotic recurrences as well as the audit of anticoagulant therapy, all within the previous 12 months (see "Patients" subsection). Every major bleeding episode or thrombosis was documented by checking the clinical notes of the patient or by medical letters if they attended other centers. We limited the study to the 12 months before the interview to reduce heterogeneity in treatments and anticoagulation surveillance and also to ensure a good recall of clinical episodes by the patients.


Patients attending our antiphospholipid clinic were enrolled consecutively from March 1 to August 31, 2000. To be included in the study, patients had to fulfill all of the following criteria:

1. Definite APS according to Sapporo criteria.9 Thus, only patients who tested positive—twice at least 6 weeks apart—for aCLs at medium or high titers and/or lupus anticoagulant (LA) entered the study.

2. History of thrombosis. Deep vein thrombosis was documented by appropriate imaging (ultrasonography and/or venography); pulmonary embolism, by high-probability lung scan, spiral computed tomographic scanning, or pulmonary arteriography; stroke, by computed tomographic scanning or magnetic resonance imaging; and peripheral and visceral arterial thrombosis, by arteriography or surgery. Cerebral transient ischemic attacks were diagnosed in the setting of acute focal neurologic symptoms or signs lasting less than 24 hours. Myocardial infarction was defined as typical chest pain with characteristic electrocardiographic features and elevated creatine kinase (MB fraction) levels.

3. Treatment with oral anticoagulants to a target INR of 3.5 (INR range, 3.0-4.0) during the previous 12 months.

Antiphospholipid antibody determinations

All the determinations were performed at the laboratories of the Lupus Research Unit (aCL assays) and the coagulation laboratories (LA assays) of St Thomas' Hospital, London, England.

The IgG and IgM aCLs were measured by β2-glycoprotein I–dependent standardized enzyme-linked immunosorbent assay, as described elsewhere.10 Results are expressed in a semiquantitative fashion as follows: negative (IgG, <5 G phospholipid [GPL] units; IgM, <3.2 M phospholipid [MPL] units); low positive (IgG, 5-15 GPL units; IgM, 3.2-6 MPL units); medium positive (IgG, 16-80 GPL units; IgM, 7-50 MPL units); and high positive (IgG, >80 GPL units; IgM, >50 MPL units).

The LA was detected according to the guidelines of the International Society on Thrombosis and Hemostasis, Scientific Subcommittee on Lupus Anticoagulant/Phospholipid-Dependent Antibodies,11 using activated partial thromboplastin time and diluted Russell viper venom time as screening tests.

Determinations of IgA aCLs or antibodies directed against other phospholipids or phospholipid-binding proteins (eg, β2-glycoprotein I–specific enzyme-linked immunosorbent assay, antiprothrombin antibodies, antiphosphatidylserine, and phosphatidylcholine) were not considered for the purposes of this study.

Audit of anticoagulant therapy

Although the target INR for every individual patient is recommended by our unit, anticoagulant monitoring is provided by their local anticoagulation clinic or general practitioner. The periodicity of INR controls was determined by the physicians responsible for anticoagulation monitoring according to their own judgment.

During the interview, every patient's personal anticoagulation book was checked, when available. The audit12 consisted of counting the number of times during the previous 12 months that the INR lay within predefined limits: less than 2.0, 2.0 to 2.9, 3.0 to 4.0, 4.1 to 4.9, 5.0 to 6.0, and more than 6.0. The proportion of measurements in each range was then calculated for the entire population. Patients who did not provide their anticoagulation book were included in the study for clinical end points (see "Clinical End Points" subsection), although they could not be included in the audit.

Clinical end points

The clinical end points were documented major bleeding and recurrent thrombosis. Only episodes of major bleeding, defined as intracranial, intraocular, gastrointestinal, retroperitoneal, or requiring transfusion or admission to a hospital, were considered for the purposes of this study. Minor bleeding was not an end point because of its clinical irrelevance and patients' difficulty in recalling such episodes. Recurrent thromboses were diagnosed by the same criteria as for original thromboses. Patients having bleeding or thrombosis were investigated for specific risk factors at the time of the event.

One of the potential problems of the study was that patients who missed the scheduled clinic could have been admitted somewhere else—or even died—because of bleeding or thrombosis, thus underestimating the rate of severe complications. However, it is usual practice for local hospitals or general practitioners looking after our patients to inform us of any complications arising. Any patients who failed to attend clinic in the study period were contacted by staff nurses to ensure they had not had any major bleeding or thrombotic events. No patient missed a clinic appointment during the study period because of major bleeding, thrombosis, or death from any cause.

Statistical analysis

Data filing and processing and statistical calculations were performed with StatView software (version 5.0.1 for Power Macintosh; SAS Institute Inc, Cary, NC). The results were expressed as rates of events per 100 patient-years, which were calculated as


In this case, the total person-years was equal to the total number of patients, as the study period was 12 months in all cases. Confidence intervals (CIs) were calculated assuming a Poisson distribution, since the results were expressed as rates, the events were random and independent, and the number of events was small, thus not allowing a normal approximation.13

Univariate comparison between continuous variables was performed by 2-tailed t test. To disclose clinical variables that could be risk factors for the specific outcomes (bleeding or thrombosis), we performed backward stepwise logistic regression, with bleeding (yes/no) and thrombosis (yes/no) used as dependent variables.


The clinical profile of our cohort is summarized in Table 1. Our study population was predominantly white, and the female-male ratio was 9:1. Half of the patients had primary APS, and most of those with the secondary form had systemic lupus erythematosus according to the 1982 American College of Rheumatology classification criteria.14 The median age was 40 years, and the median time receiving anticoagulation was 5 years. Twenty-three patients (35%) had been taking warfarin for less than 5 years and 11 (17%) for less than 3 years. Most patients (51 [77%]) had a history of arterial thrombosis, and 38 patients (58%) had a history of stroke. The prevalence of the different aPLs is shown in Table 2.

Audit of anticoagulant therapy

Data on anticoagulation surveillance were available for 59 patients (89%). One thousand fifty INR determinations were performed in a total of 638 patient-months, thus resulting in 1.65 determinations per patient-month. The range of determinations per patient during the study period was 5 to 43.

The cumulative number of INR determinations within each predefined range is shown in Figure 1. Of the INRs, 391 (37%) were between 3.0 and 4.0, 323 (31%) between 2.0 and 2.9, and 137 (13%) between 4.0 and 4.9. At the extremes, 123 (12%) were lower than 2.0 and 77 (7%) were higher than 5.0.


Bleeding rates are given in Table 3, and details regarding bleeding episodes are given in Table 4.

Major bleeding was observed in 4 patients (2 retroperitoneal, 1 intracranial, and 1 rectal), resulting in a rate of 6 cases per 100 patient-years (95% CI, 1.6-15.0). The rate of intracranial bleeding was 1.5 per 100 patient-years (95% CI, 0.04-8.4). None of the bleeding episodes was fatal. In a population of 66 patients, this results in a 95% CI of 0 to 3.7 deaths due to bleeding per 100 patient-years.

One patient had bilateral subdural hematomas in South America, when she omitted anticoagulation monitoring for more than 1 month. At the time of bleeding, she had experienced a recent head trauma and her INR was more than 20. After surgical intervention, she had full recovery. Two patients had retroperitoneal bleeding when they restarted warfarin treatment after a transient withdrawal for a renal biopsy. The INRs at the time of the events were 5.2 and 1.9. Finally, 1 patient suffered continuous rectal bleeding at INRs below 3 until surgery for internal hemorrhoids was performed. Thus, all patients who bled had a clear precipitating factor. In addition, 2 patients were taking aspirin (Table 4).


Details of thrombotic episodes are summarized in Table 4. Six patients had thrombotic recurrences during the follow-up, a rate of 9.1 per 100 patient-years (95% CI, 3.3-19.6). Four episodes were arterial and 2 were venous thromboses. Three patients with arterial events had additional risk factors such as high blood pressure and cigarette smoking. The woman who suffered subdural hematomas had a subsequent venous thrombosis when deprived of thromboprophylaxis postoperatively. Two patients had no other risk factors for thrombosis. The INRs at the time of thrombosis were between 2.1 and 2.6 (it was not determined in 1 case). No patient died of thrombosis (95% CI, 0-3.7). Five of 6 patients experienced a recurrence in the same vascular bed as the original thrombosis.

Predictors of events

The following clinical variables were included in 2 logistic regression models to determine which could predict bleeding or thrombosis: age, time receiving anticoagulant therapy, primary vs secondary APS, aCL positive, LA positive, previous arterial thrombosis, previous stroke, previous venous thrombosis, and previous thrombocytopenia. Sex and race were not included in the model, since more than 90% of the cohort were women and white.

None of the studied variables predicted bleeding. However, each successive year of anticoagulation treatment increased the odds of thrombotic recurrences within the previous year by 1.25 (95% CI, 1.03-1.5). Consistently, patients with recurrent thrombosis within the previous year had been taking anticoagulants for a mean of 11 years, while those without recurrent events had taken warfarin for a mean of 6 years (mean difference, 5.0 years; 95% CI, 1.4-8.7 years; P = .008, 2-tailed t test).


Since 1995, we have adopted oral anticoagulation to a target INR of 3.5 (range, 3.0-4.0) as the standard therapy for patients with aPLs and previous thromboses. In this study, we found that this practice does not result in a high incidence of intracranial or fatal bleeding in patients with definite APS who are at a high risk of recurrent thrombosis.

The actual risk of thrombosis in patients with aPLs is not well established, particularly for patients without previous events.15 We know that the presence of LA or aCLs at medium to high titers increases that risk, especially in the setting of a history of thrombosis.16,17 In retrospective series1-3 the rates of thrombotic recurrences in untreated subjects were between 19 and 29 events per 100 patient-years.

The risk of bleeding during oral anticoagulant treatment is not well defined in the general population. Table 5 summarizes the results of 5 relevant studies that cover this issue: a meta-analysis on the efficacy of anticoagulation in atrial fibrillation18; a randomized controlled trial of anticoagulation vs aspirin after a transient ischemic attack19; a prospective cohort of anticoagulation in clinical practice20; a prospective inception cohort to study bleeding complications of oral anticoagulant therapy21; and a randomized controlled trial of long vs short anticoagulation in venous thromboembolism.22

In the meta-analysis by Hart et al,18 the risk per year of major extracranial bleeding and intracranial hemorrhage was 0.9% and 0.3%, respectively. These figures are in keeping with the results of the Italian prospective inception cohort21 and the prospective series of Kalra et al.20 However, other authors have found higher rates of bleeding.19,22 The results of the Stroke Prevention in Reversible Ischemia Trial (SPIRIT) were particularly striking. In that study, patients with a recent transient ischemic attack allocated to receive oral anticoagulants to an INR of between 3.0 and 4.5 suffered major and intracranial bleeding at very high rates of 7 and 3.7 per 100 patient-years, respectively.19

Factors that may increase the risk of hemorrhage during oral anticoagulant therapy include age greater than 65 to 75 years,21,23-26 receiving multiple medications,26 history of gastrointestinal tract bleeding,23,25 arterial thrombosis21 or stroke,25 and INRs higher than 4.5 to 5.0.12,21 If one focuses specifically on intracerebral bleeding, the presence of leukoariosis (diffuse white matter hypodensity on magnetic resonance imaging or computed tomographic scanning) has been recognized as the most important risk factor for bleeding in the SPIRIT cohort.26 In the same study, the intensity of anticoagulation was not an independent predictor of bleeding in the multivariate model.26

In our group, anticoagulation control was not as strict as we wished. The INRs within the "therapeutic" range between 3.0 and 4.0 were accomplished in only 37% of determinations. Although the sort of audit we used implied a bias toward an increase in "out of range" determinations—patients with good control were tested less often—our results may also reflect the fears of many anticoagulation clinics of long-term high-intensity anticoagulation in routine clinical practice. In the United Kingdom, the majority of patients attending anticoagulant clinics are receiving warfarin because they have atrial fibrillation and are aiming for a target INR of 2.0 to 3.0. Our policy did, however, maintain INRs above 2.0 in almost 90% of the cases, while only 7% of determinations were above 5.0. When a target INR of 2.5 (range, 2.0-3.0) is used, subtherapeutic INRs below 2.0 are found much more frequently,12,20 a fact that may have devastating implications in patients at high risk for thrombotic recurrences.

In fact, our population was at a particularly high risk for recurrent thrombosis: we found 9 thrombotic events per 100 patient-years during anticoagulant therapy, higher than the risk observed in the studies not including patients with APS (Table 5). This rate was also higher than that in patients receiving anticoagulation at INRs higher than 3.0 from retrospective series,1,2 thus highlighting the different risk of recurrences depending on aCL titers and presence of LA.16,17 The tendency of patients with APS to have recurrent thrombosis in the same vascular bed (venous or arterial) was notable in our group, as in other series.17,27 In this study all thrombotic events happened at INRs between 2.1 and 2.6. Nevertheless, 66% of patients with thrombosis during the previous year had associated risk factors (such as smoking or hypertension), a fact that has been observed by other authors.3

We observed a risk of major bleeding that was intermediate between those of the SPIRIT study and groups of patients with atrial fibrillation (Table 5). In particular, our frequency of intracranial bleeding was within the 95% CI of the "clinical practice" cohort of Kalra et al20 and was not substantially higher than in other series (Table 5). The increased risk of using a high target INR may have been compensated for by the fact that patients with APS are generally young, since advanced age is a consistent risk factor for cerebral bleeding in patients receiving anticoagulation.21,23-26 Moreover, the differences seen in the frequency of intracerebral bleeding between our series and that of Gorter26 may also suggest that the nature of the presumed small vessel damage may not be the same in patients with cerebral thrombosis caused by APS and that caused by long-standing arterial hypertension and/or diabetes. Of note, in a recent article from our unit that describes 15 cases of major bleeding in patients with APS between 1989 and 1999, we did not find any episode of intracerebral bleeding.28 Finally, concurrent risk factors, including renal biopsies in 2 patients, were present in all patients who bled in our series.

Probably in part because of sample size restrictions, we could not define factors that predict bleeding or recurrent thromboses in patients with APS. The only relevant finding in the multivariate analysis was that the risk of thrombotic recurrences does not decrease—it actually increases—with time. Therefore, this study supports the belief that indefinite anticoagulation is indicated for this group.1,2,4,22

Our study has obvious limitations. The main one is its retrospective design. This was partially compensated for by personally interviewing every patient, which allowed recognition of major bleeding and thrombotic episodes even if treated at other hospitals. The audit of anticoagulant therapy did not calculate the actual time at each range of INRs, the most accurate method of monitoring.20 Our audit does, however, offer an indication of the quality of anticoagulation in our patients in the United Kingdom. Finally, the relatively small sample size widened CIs.

On the other hand, to our knowledge, this is the first study of anticoagulation including patients who fulfilled Sapporo criteria. Thus, our population had definite APS (ie, low-titer aCLs were not included) and was at high risk of thrombotic recurrences, as our own results confirm. Arterial events, which are more disabling than venous, were also prevalent. All this results in a critical difference from other series labeled APS,5 which reported a much more benign course of the disease and whose conclusions have been the support for recent guidelines recommending a target INR of 2.5 for all patients with APS treated with oral anticoagulants.8

When the indication for and intensity of anticoagulant therapy are determined in an individual patient, bleeding risk is clearly not the only factor to consider. It must be weighed against the possibility of recurrent serious thrombosis.23,29 On the basis of a cautious analysis of our results and those of previous studies,1,2 we believe that most patients with definite APS and previous thrombosis should be treated to a target INR of 3.5. The exception could be individuals with only venous events and those at high risk of bleeding (aged or with previous life-threatening bleeding episodes), who could be considered for lower intensities of anticoagulation.29 Every effort must be made to limit concomitant medications and invasive procedures in this group. Furthermore, the adequate control of additional risk factors for thrombosis, such as smoking, hypertension, and hyperlipidemia, must be considered an integral part of therapy in patients with definite APS.

Accepted for publication September 25, 2001.

This study was supported by grant 99/5007 of Fondo de Investigacion Sanitaria, Spain; the Department of Health of the Basque Government, Vitoria-Gasteiz, the Basque Country, Spain; and Lupus UK, Essex, England.

Corresponding author and reprints: Munther A. Khamashta, MD, MRCP, PhD, Lupus Research Unit, The Rayne Institute, St Thomas' Hospital, London SE1 7EH, England (e-mail:

Rosove  MHBrewer  PMC Antiphospholipid thrombosis: clinical course after the first thrombotic event in 70 patients.  Ann Intern Med. 1992;117303- 308Google ScholarCrossref
Khamashta  MACuadrado  MJMujic  FTaub  NAHunt  BJHughes  GRV The management of thrombosis in the antiphospholipid-antibody syndrome.  N Engl J Med. 1995;332993- 997Google ScholarCrossref
Krnic-Barrie  SO'Connor  CRLooney  SWPierangeli  SSHarris  EN A retrospective review of 61 patients with antiphospholipid syndrome: analysis of factors influencing recurrent thrombosis.  Arch Intern Med. 1997;1572101- 2108Google ScholarCrossref
Derksen  RHWMde Groot  PGKater  LNieuwenhuis  HK Patients with antiphospholipid antibodies and venous thrombosis should receive long term anticoagulant treatment.  Ann Rheum Dis. 1993;52689- 692Google ScholarCrossref
Schulman  SSvenungsson  EGranqvist  Sand the Duration of Anticoagulation Study Group, Anticardiolipin antibodies predict early recurrence in thromboembolism and death among patients with venous thromboembolism following anticoagulant therapy.  Am J Med. 1998;104332- 338Google ScholarCrossref
Ginsberg  JSWells  PSBrill-Edwards  P  et al.  Antiphospholipid antibodies and venous thromboembolism.  Blood. 1995;863685- 3691Google Scholar
British Committee for Standards in Haematology, Guidelines on oral anticoagulation: third edition.  Br J Haematol. 1998;101374- 387Google ScholarCrossref
Greaves  MCohen  HMachin  SJMackie  I Guidelines on the investigation and management of the antiphospholipid syndrome.  Br J Haematol. 2000;109704- 715Google ScholarCrossref
Wilson  WAGharavi  AEKoike  T  et al.  International consensus statement on preliminary classification criteria for definite antiphospholipid syndrome.  Arthritis Rheum. 1999;421309- 1311Google ScholarCrossref
Harris  ENGharavi  AEPatel  SPHughes  GRV Evaluation of the anticardiolipin antibody test: report of a standardization workshop, held April 4th, 1986.  Clin Exp Immunol. 1987;68215- 222Google Scholar
Brandt  JTTriplett  DAAlving  BScharrer  I Criteria for the diagnosis of lupus anticoagulants: an update.  Thromb Haemost. 1995;741185- 1190Google Scholar
Rose  PE Audit of anticoagulant therapy.  J Clin Pathol. 1996;495- 9Google ScholarCrossref
Kirkwood  BR Essentials of Medical Statistics.  Oxford, England Blackwell Science Ltd1988;
Tan  EMCohen  ASFries  JF  et al.  The 1982 revised criteria for the classification of systemic lupus erythematosus.  Arthritis Rheum. 1982;251271- 1277Google ScholarCrossref
Khamashta  MA Primary prevention of thrombosis in subjects with positive antiphospholipid antibodies.  J Autoimmun. 2000;15249- 253Google ScholarCrossref
Muñoz-Rodriguez  FFont  JCervera  R  et al.  Clinical study of 100 patients with the antiphospholipid syndrome.  Semin Arthritis Rheum. 1999;29182- 190Google ScholarCrossref
Finazzi  GBrancaccio  VMoia  M  et al.  Natural history and risk factors for thrombosis in 360 patients with antiphospholipid antibodies: a four-year prospective study from the Italian registry.  Am J Med. 1996;100530- 536Google ScholarCrossref
Hart  RGBenavente  OMcBride  RPearce  LA Antithrombotic therapy to prevent stroke in patients with atrial fibrillation: a meta-analysis.  Ann Intern Med. 1999;131492- 501Google ScholarCrossref
The Stroke Prevention in Reversible Ischemia Trial (SPIRIT) Study Group, A randomized trial of anticoagulants versus aspirin after cerebral ischemia of presumed arterial origin.  Ann Neurol. 1997;42857- 865Google ScholarCrossref
Kalra  LYu  GPerez  ILakhani  ADonaldson  N Prospective study to determine if trial efficacy of anticoagulation for stroke prevention in atrial fibrillation translates into clinical effectiveness.  BMJ. 2000;3201236- 1239Google ScholarCrossref
Palareti  GLeali  NCoccheri  S  et al.  Bleeding complications of oral anticoagulant treatment: an inception-cohort, prospective collaborative study (ISCOAT).  Lancet. 1996;348423- 428Google ScholarCrossref
Kearon  CGent  MHirsh  J  et al.  A comparison of three months of anticoagulation with extended anticoagulation for a first episode of idiopathic venous thromboembolism.  N Engl J Med. 1999;340901- 907Google ScholarCrossref
Levine  MNRaskob  GLandefeld  SKearon  C Hemorrhagic complications of anticoagulant treatment.  Chest. 2001;119 ((suppl)) 108S- 121SGoogle ScholarCrossref
The Stroke Prevention in Atrial Fibrillation Investigators, Bleeding during antithrombotic therapy in patients with atrial fibrillation.  Arch Intern Med. 1996;156409- 416Google ScholarCrossref
Beyth  RJQuinn  LLandefeld  CS A multicomponent intervention to prevent major bleeding complications in older patients receiving warfarin: a randomized, controlled trial.  Ann Intern Med. 2000;133687- 695Google ScholarCrossref
Gorter  JW Major bleeding during anticoagulation after cerebral ischemia: patterns and risk factors.  Neurology. 1999;531319- 1327Google ScholarCrossref
Tektonidou  MGIoannidis  JPABoki  KAVlachoyiannopoulos  PGMoutsopoulos  HM Prognostic factors and clustering of serious clinical outcomes in antiphospholipid syndrome.  QJM. 2000;93523- 530Google ScholarCrossref
Castellino  GCuadrado  MJGodfrey  TKhamashta  MAHughes  GRV Characteristics of patients with antiphospholipid syndrome with major bleeding after oral anticoagulant treatment.  Ann Rheum Dis. 2001;60527- 530Google ScholarCrossref
Ruiz-Irastorza  GKhamashta  MAHughes  GRV Antiaggregant and anticoagulant therapy in systemic lupus erythematosus and Hughes' syndrome.  Lupus. 2001;10241- 245Google ScholarCrossref