|Home | About | Journals | Submit | Contact Us | Français|
OBJECTIVE: To determine the strength of evidence supporting an accentuated bleeding risk when patients with CHADS2 risk factors (chronic heart failure, hypertension, advanced age, diabetes, and prior stroke/transient ischemic attack) receive warfarin.
METHODS: A systematic literature search of MEDLINE (January 1, 1950, through December 22, 2009) and Cochrane CENTRAL (through December 22, 2009) was conducted to identify studies that reported multivariate results on the association between CHADS2 covariates and risk of bleeding in patients receiving warfarin. Each covariate was evaluated for its association with a specific type of bleeding. Individual evaluations were rated as good, fair, or poor using methods consistent with those recommended by the Agency for Healthcare Research and Quality. The strength of the associations between each CHADS2 covariate and a specific type of bleeding was determined using Grading of Recommendations Assessment, Development and Evaluation criteria as insufficient, very low, low, moderate, or high for the entire body of evidence.
RESULTS: Forty-one studies were identified, reporting 127 multivariate evaluations of the association between a CHADS2 covariate and bleeding risk. No CHADS2 covariate had a high strength of evidence for association with any bleeding type. For the vast majority of evaluations, the strength of evidence between covariates and bleeding was low. Advanced age was the only covariate that had a moderate strength of evidence for association; this was the strongest independent positive predictor for major bleeding. Similar findings were observed regardless of whether all included studies, or only those evaluating patients with atrial fibrillation, were assessed.
CONCLUSION: The associations between CHADS2 covariates and increased bleeding risk were weak, with the exception of age. Given the known association of the CHADS2 score and stroke risk, the decision to prescribe warfarin should be driven more by patients' risk of stroke than by the risk of bleeding.
AF = atrial fibrillation; AHR = adjusted hazard ratio; AOR = adjusted odds ratio; CHADS2 = chronic heart failure, hypertension, advanced age, diabetes, and prior stroke/transient ischemic attack; CHF = chronic heart failure; CI = confidence interval; CVD = cerebrovascular disease; HTN = hypertension; TIA = transient ischemic attack
Approximately 5% of patients with atrial fibrillation (AF) will develop an embolic stroke each year.1,2 However, this risk is not equally distributed among patients. The CHADS2 scoring system (in which 1 point is assigned for chronic heart failure [CHF], hypertension [HTN], advanced age, and diabetes and 2 points for prior stroke/transient ischemic attack [TIA]) was derived in an effort to predict cardioembolic stroke risk based on the presence of 1 or more risk factors in a large population of untreated patients with AF. In large clinical trials, warfarin therapy reduced the risk of embolic stroke vs aspirin or placebo in patients with AF and 2 or more CHADS2 risk factors.1-6 However, a recent meta-analysis found that less than half of patients with AF and an indication for therapy received warfarin (48%; 95% confidence interval [CI], 43%-54%).7
Several studies suggest that physicians withhold warfarin therapy for many patients because of a perceived increased risk of bleeding.8-10 Like the risk of embolic stroke, the occurrence of bleeding may be related to individual patient characteristics. Because the CHADS2 risk factors for the development of embolic stroke in patients with AF are also listed in the approved warfarin-prescribing information as risk factors for bleeding, clinicians are caught in a quandary.11 A direct evaluation of the evidence supporting an association between CHADS2 risk factors and bleeding would assist clinicians in weighing the potential benefit and detriment of anticoagulant prophylaxis for their patients with AF.
Therefore, we conducted a systematic review of evidence to determine the strength of the association between the components of CHADS2 and bleeding risk in patients treated with warfarin, first irrespective of indication and then focusing on only those with AF.
Two investigators conducted a systematic literature search independently using MEDLINE (January 1, 1950, through December 22, 2009) and Cochrane CENTRAL (through December 22, 2009). The complete search strategy is available in the Supporting Online Material (a link to which is provided at the end of this article). A manual review of references from each pertinent article, identified review articles, and treatment guidelines was also conducted to identify additional articles.
Studies were eligible for inclusion in the systematic review if they: (1) reported on a population of patients receiving warfarin, (2) reported on the association between CHADS2 characteristics and the risk of any bleeding event, (3) conducted multivariate analysis to determine the association between patient characteristics and risk of bleeding and reported results on at least 1 covariate of interest, and (4) were published in the English language. Two investigators (W.T.C. and V.T.) determined study eligibility independently, with disagreements resolved by discussion or by a third investigator (C.I.C.).
For each included study, 2 independent investigators (W.T.C. and V.T.) abstracted data on the following information: author, year, study design, population included, indication for warfarin, duration of follow-up in patient-years, previous warfarin use, country, “major” or “minor” or “any” bleeding definition, event rate (as bleeding events per 100 patient-years), P value for the univariate association between a patient characteristic (covariate) and bleeding, and effect size and P value for the multivariate association between a covariate and bleeding. Covariates of interest included CHF, HTN, advanced age, diabetes, and prior stroke/TIA (cerebrovascular disease [CVD]) (or the CHADS2 risk factors). Qualitative synthesis of data is reported using descriptive statistics.
In this systematic review, an evaluation was defined as an assessment of a covariate for its association with a specific type of bleeding. Each evaluation was rated for its validity using methods consistent with those recommended by the Agency for Healthcare Research and Quality.12 More specifically, hierarchy of study design, objectivity of bleeding definition, sample size, overall event rate, and magnitude of effect size were considered. Individual evaluations were then given an overall ranking of good, fair, or poor.
We used the criteria and methods of GRADE (Grading of Recommendations Assessment, Development, and Evaluation) to assess the strength of the body of evidence for each CHADS2 covariate. This system uses 4 required domains: risk of bias, consistency, directness, and precision.13 Two investigators (O.J.P. and D.M.S.) made all assessments (with disagreements resolved through discussion). The evidence pertaining to each of the CHADS2 covariates was classified into 5 broad categories: high, moderate, low, very low, or insufficient. The features that determined the strength of evidence for the different outcomes evaluated in this review are described in more detail in Table S1 (Supporting Online Material).
Although our base-case analyses evaluated patients with any indication for oral anticoagulation receiving warfarin therapy, in subgroup analysis we limited eligible studies to those enrolling patients who had AF and were receiving warfarin therapy.
Our literature search revealed 3324 nonduplicate citations. On title and abstract review, 923 citations were excluded, leaving 2401 citations for full-text review. On full-text review, 2360 were excluded (Figure 1). A total of 41 studies published between 1988 and 2009 from 9 different countries were included in this systematic review.14-55 These studies included 2 post hoc analyses of randomized controlled trials, 12 prospective observational studies, 22 retrospective observational studies, 4 mixed prospective and retrospective studies, and 1 study of unclear study design.
Key characteristics of included studies can be found in Table 1. The total sample size was 166,871, with the number of patients enrolled in a study ranging from 66 to 6988. Total duration of follow-up ranged from 66 to 133,976 patient-years (median, 1163 patient-years). Indications for warfarin therapy consisted of AF, venous thromboembolism prophylaxis and treatment, CVD, myocardial infarction, valve replacement, and antiphospholipid syndrome. Forty-one percent of studies did not report previous warfarin use (an inception cohort), and the remaining studies included patients who began receiving warfarin therapy during the study or had some previous warfarin use.
All 41 studies reported results of multivariate analysis as either adjusted odds ratios (AORs) or adjusted hazard ratios (AHRs). Among the 41 studies, 168 evaluations assessed the association between a CHADS2 covariate and bleeding (either univariate or multivariate), with all studies reporting at least one multivariate result and 127 evaluations (76%) providing multivariate results. Although 39 (31%) of the 127 multivariate evaluations reported a significant association between the CHADS2 covariates and any types of bleeding, the remaining 88 (69%) found no significant association (Table 2). Of the evaluations, 72% were for major bleeding, with the remaining split between minor (11%) and any bleeding (17%) (see Tables S2-S7 in Supporting Online Material).
Thirteen studies (n=16,333) reported 22 evaluations of the association between CHF and bleeding.14-26 Of the evaluations, 5%, 59%, and 36% were rated as good, fair, or poor, respectively. The prevalence of CHF in the studies ranged from 5% to 76%. Major bleeding events ranged from 0.8 to 10.8 events per 100 patient-years, whereas none reported the event rates for minor bleeding. Only 1 evaluation reported the event rate of any bleeding as 6.2 events per 100 patient-years.
Thirteen (59%) of the 22 evaluations included multivariate results, of which 7 (54%) (6 fair, 1 poor quality), 3 (23%), and 3 (23%) were for major (Figure 2), minor, and any bleeding, respectively. Among these, only 1 evaluation found CHF to be an independent positive predictor of major bleeding (Figure 2 and Table S2 in Supporting Online Material).25 No significant associations between CHF and minor or any bleeding were seen (Table 2). We graded the strength of the body of evidence for this association as low for major, minor, and any bleeding.
Twenty-one studies (n=88,151) reported 33 evaluations of the association between HTN and bleeding.14-25,27-35 Of the evaluations, 12%, 46%, and 42% were rated as good, fair, and poor, respectively. The prevalence of HTN in studies ranged from 4% to 79%. The event rates of major and any bleeding ranged from 0.8 to 10.8 and from 3.7 to 12.2 events per 100 patient-years, respectively; event rates for minor bleeding were not reported.
Of the 33 evaluations, 19 (58%) included multivariate results, of which 12 (63%) (1 good, 6 fair, 5 poor quality), 3 (16%), and 4 (21%) were for major (Figure 2), minor, and any bleeding, respectively. Six evaluations (32%), 2 from each bleeding type (for 1 minor bleeding association, statistical significance was suggested because the 95% CIs did not cross the line of unity; P=.06), found HTN to be an independent positive predictor of any type of bleeding, with an AOR of bleeding ranging from 2.3 to 3.6 and an AHR of 1.10 to 1.25 (see Table S3 in Supporting Online Material). The remaining evaluations found no significant association between HTN and either minor or any bleeding. We graded the strength of the body of evidence for this association as low for major, minor, and any bleeding (Table 2).
Thirty-eight studies (n=165,226) reported 59 evaluations of the association between age and bleeding.14-17,19-22,24,25,27-55 Of the evaluations, 20%, 51%, and 29% were rated as good, fair, and poor, respectively. Age was evaluated as both a continuous variable (linear age per year, per 5 years or per decade) and categorical variable (with the most common cutoffs at 65 years, 75 years, 80 years, or 90 years). Major, minor, and any bleeding event rates ranged from 0.5 to 10.8, 2.7 to 21.8, and 3.7 to 13.0 events per 100 patient-years, respectively.
Of the 59 evaluations, 55 (93%) included multivariate results, of which 37 (67%) (7 good, 23 fair, 7 poor quality), 6 (11%), and 12 (22%) were for major (Figure 2), minor, and any bleeding, respectively. Of the 22 evaluations (40%), all rated as good or fair quality, that found age to be an independent positive predictor of bleeding, 18 (82%) were for major bleeding. Effect sizes for significant major bleeding risks were as high as an AOR of 2.5 when evaluating age as a linear variable per 5 years17 and as high as an AHR of 2.75 when evaluating patients older than 75 years vs those younger than 75 years (see Table S4 in Supporting Online Material).21 Thus, evidence of moderate strength suggests that age increases major bleeding risk in those receiving warfarin. Because less than 40% of the multivariate evaluations found either a significant or nonsignificant positive association between advanced age and minor or any bleeding risk in those receiving warfarin, the strength of evidence suggesting that age increases these bleeding types was low (Table 2).
Eighteen studies (n=106,336) reported 29 evaluations of the association between diabetes and bleeding.14-19,21-25,27,29,31,32,34-36 Of the 29 evaluations, 7%, 55%, and 38% were rated as good, fair, and poor, respectively. The prevalence of diabetes in the studies ranged from 4% to 53%. Major bleeding events ranged from 0.8 to 10.8 events per 100 patient-years; event rates for minor bleeding were not reported. Only 1 evaluation reported the event rate of any bleeding as 7.7 events per 100 patient-years.
Of the 29 evaluations, 19 (66%) reported multivariate results, of which 13 (68%) (2 good, 8 fair, 3 poor quality), 3 (16%), and 3 (16%) were for major (Figure 2), minor, and any bleeding, respectively. Only 3 multivariate evaluations of good or fair quality (16%), 2 for major bleeding and 1 for any bleeding, found diabetes mellitus to be an independent positive predictor of bleeding. In 1 study, the AOR for major bleeding in patients with diabetes was 4.4 (see Table S5 in Supporting Online Material).23 No significant associations between diabetes and minor bleeding were seen. We graded the strength of evidence for this association as low for major or any bleeding and very low for minor bleeding (Table 2).
Thirteen studies (n=93,020) reported 25 evaluations of the association between CVD and bleeding.15-17,19,20,23,24,27,30,32,33,51,55 Of the evaluations, 4%, 60%, and 36% were rated as good, fair, and poor, respectively. The definition of CVD varied by study, ranging from history of stroke, TIA, or cerebrovascular stroke. The prevalence of CVD in studies ranged from 1% to 58%. The event rates of major and any bleeding ranged from 0.6 to 8.1 and from 3.7 to 12.2 events per 100 patient-years; the event rates for minor bleeding were not reported.
Of the 25 evaluations, 21 (84%) included multivariate results, of which 14 (67%) (1 good, 8 fair, 5 poor), 2 (10%), and 5 (24%) were for major (Figure 2), minor, and any bleeding, respectively. A total of 7 multivariate evaluations (33%) found CVD to be an independent predictor of bleeding (6 for major bleeding and 1 for any bleeding). The AOR of major bleeding for those with CVD was as high as 3.6, and the AHR of any bleeding was 1.12 in patients with previous stroke (see Table S6 in Supporting Online Material).23,27 We graded the strength of evidence for this association as low for major and any bleeding and very low for minor bleeding (Table 2).
Only 1 study of fair quality (n=783) evaluated the association between CHADS2 score, reported as a linear variable, and any bleeding.33 The bleeding event rate in the study was 3.7 events per 100 patient-years. Multivariate results were not reported. Thus, data are insufficient to grade the strength of evidence for the association between the CHADS2 score and any bleeding risk (Table 2).
Thirteen studies (n=50,448) evaluated an association between a CHADS2 covariate in a population in which all patients had AF and were receiving warfarin therapy.15,17,18,23-25,33,34,36,38-40,43,54 Follow-up ranged from 355 to 18,946 patient-years (median, 814 patient-years). A total of 69 evaluations reported an association between a covariate and bleeding, of which 52 (75%) were multivariate (30, 10, and 12 evaluations for major, minor, and any bleeding, respectively).
Among the 30 multivariate evaluations for major bleeding, more evaluations (n=10) assessed the association between age and major bleeding in patients with AF who were receiving warfarin, as compared with those with CHF (n=4), HTN (n=4), diabetes (n=8), or CVD (n=4). Of the 10 evaluations (n=39,672) reporting a multivariate association between age and major bleeding, 9 (90%) found a positive association with major bleeding, 5 (50%) of which were significant. We graded the strength of evidence for this association to be moderate. Less than half of the multivariate results showed CHF, diabetes, and CVD to be independent positive predictors for major bleeding, with a low strength of evidence. No significant association between HTN and major bleeding was noted, with the strength of evidence graded as very low (Table 3).
Of the 10 multivariate evaluations for minor bleeding, 2 (20%) showed HTN to have a significant positive association with minor bleeding. None of the other covariates (CHF, age, diabetes, or CVD) were found to be significant predictors of minor bleeding. We graded the strength of evidence for HTN to be low, with the other covariates being very low.
Of the 12 multivariate evaluations for any bleeding, one found HTN and another found age to be independent positive predictors of any bleeding. None of the other evaluations found CHF, advanced age, diabetes, or CVD to have any effect on any bleeding. We graded the strength of evidence for each of these factors to be very low.
Overall, associations identified from studies enrolling only patients with AF were rated to be of similar quality and provided comparable conclusions regarding strength of evidence as those of the overall analysis (Figure 2, Tables Tables22 and and33).
The number of patients and patient evaluations in our study was generally greater for major bleeding than minor or any bleeding. None of the CHADS2 covariates had a high strength of association with warfarin-related bleeding risk. The strength of evidence of most covariates was low to very low, with the exception of advanced age and major bleeding. The moderate strength of association between age and major bleeding risk was the strongest among all covariates, whereas the very low strength of association between diabetes and minor bleeding was the weakest. The subgroup analysis in our study revealed that advanced age was also the strongest independent positive predictor of bleeding risk among all CHADS2 covariates in patients specifically with AF. As in the full analysis, the strength of associations between the other covariates and bleeding risk was very low.
Diener et al38 and Lip et al56 published an evaluation comparing the predictive value of 5 different contemporary bleeding risk stratification schemas using combined SPORTIF III and V warfarin data (N=3665). Of the schemas evaluated, the HAS-BLED (Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol Concomitantly) score appeared to be the best predictor of bleeding in patients taking warfarin (C statistic, 0.66; 95% CI, 0.61-0.70); however, all schemas provided only moderate predictive ability (C statistic, 0.52-0.66; a C statistic of 0.50 implies that a schema is no better at predicting a bleeding event than random guessing). The fact that no one schema was vastly superior to another despite each using different risk factors for bleeding is not surprising given the current results of our systematic literature review. The preponderance of the medical literature suggests that many risk factors have either conflicting or scant data supporting their association with bleeding of any severity, clarifying why developing a highly predictive schema has proved difficult. Of the 5 schemas, none used heart failure as a risk factor, 2 (40%) used HTN, 3 (60%) used stroke, and 1 used diabetes by itself (20%). In contrast, age was used as a risk factor in all 5 schemas, albeit in different forms (continuous and categorical formats with different cutoffs for advancing age). The use of advancing age to predict bleeding is very much supported by the findings of our systematic review, which suggest that age is a relatively reliable risk factor for bleeding.
In the evaluation of CHADS2 covariates, it is important to recognize that one covariate may be serving as a surrogate for a different but related risk factor. Therefore, we caution that results must not be taken at face value. For example, the association seen with increasing age and bleeding risk may also be attributed to a decline in renal function, multiple comorbid conditions, or concomitant use of many other drugs. Furthermore, why these risk factors are associated with bleeding risk is often unknown.
Unlike the weak association with bleeding risk, the use of CHADS2 covariates in predicting stroke risk is relatively well established.56,57 Previous studies found that the risk of stroke in patients with AF increased as the number of CHADS2 risk factors increased, with an increase of up to 1.5 times for each 1-point increase in the CHADS2 score when no anticoagulant therapy was given. Warfarin therapy significantly reduced stroke risk, by approximately two-thirds, from 4.5% to 1.4%, when compared with placebo or no treatment, and by two-fifths when compared with aspirin.2-5,56 The Anticoagulation and Risk Factors in Atrial Fibrillation cohort study showed that the highest net benefit of warfarin was among patients with moderate to high risk for stroke because the absolute increase in risk for intracranial hemorrhage due to warfarin remains fairly stable across CHADS2 risk categories.58
Several limitations must be considered when interpreting the results of our systematic review. Included studies defined bleeding end points in various ways. Although we limited ourselves to descriptive analysis only (no pooling of results through meta-analysis) and stratified our analyses by bleeding severity (major, minor, or any), important residual heterogeneity may still have been present.
Our systematic review included mostly observational studies, which have inherent selection and information biases that may result in an erroneous estimate of association. Selection bias may have occurred if patients were included in the selected studies because risk factors were presumed to be associated with bleeding. Furthermore, information bias could have occurred if patients were misclassified as having or not having a potential risk factor (ie, HTN, diabetes) or as having or not having a bleeding event. Observational studies can only reveal potential associations between covariates and the risk of bleeding; they cannot prove causality.
Negative reporting bias is a potential limitation of any systematic review. Authors commonly omit insignificant results from their publications, resulting in conclusions biased toward a covariate being a risk factor for bleeding. Of the 41 studies included in our systematic review, 27 (66%) reported insignificant results, suggesting that this is of some potential concern in this systematic review. However, this bias would tend to further weaken the associations we have found. In addition, many of the studies included in this systematic review were underpowered (type 2 error) because the presence and/or occurrence of covariate or bleeding events was low in the patient samples. For example, in the study by Landefeld and Goldman,20 the failure to find CVD as a risk factor for major bleeding may be explained by the low baseline presence of CVD (only 4 of 375 patients) in the overall population. We devised an analysis protocol to accommodate for this, a description of which follows.
Given these cautions, limitations, and caveats, we used a strength-of-evidence rating scale that incorporated many features, allowing decision makers to draw conclusions on the basis of the preponderance of the evidence and reducing the influence of any 1 potentially biased study on the process. First, we evaluated the number of evaluations and patients included because many evaluations allow for the assessment of consistency, and larger evaluations are less likely to be underpowered. We then evaluated both the number and corresponding percentage of the total number of evaluations for a covariate when there was a significant positive association or a number of evaluations with a positive direction of effect regardless of significance. This gave us insight into the effect of a covariate on bleeding and the consistency of the effect across studies. Finally, we evaluated the quality of the evaluations because higher-quality evidence increases confidence in the results.
In light of their expected efficacy, safety, and ease of use, the newer oral anticoagulants (eg, Factor Xa inhibitors, direct thrombin inhibitors) are likely to replace warfarin therapy in many patients. Because the risk of bleeding associated with various patient characteristics is not yet known specifically with these agents, new data will be required. Fortunately, detailed analysis of the large-scale phase 3 clinical trials of these agents should provide much stronger evidence than is available for warfarin.
This systematic review found that the strength of evidence supporting the association between CHADS2 covariates and bleeding risk is low to very low, with the exception of advanced age, for which the strength of evidence is moderate. This appeared true regardless of whether the assessment included all studies or only those evaluating patients with AF. It is important to note that these associations were not as strong as those between the CHADS2 covariates and stroke risk. Thus, these findings support the recommendation that the decision to prescribe oral anticoagulant prophylaxis to patients with AF should be driven more by patients' risk of stroke than by the risk of bleeding due to these factors.
We would like to thank Dr Kearon and colleagues for providing additional data pertaining to their trial (Kearon et al47); Erica L. Baker, PharmD, and Devi Mathur, MD, for their help with study procurement and data collection; and Ruth Sussman, PhD, for her editorial assistance.
Supporting Online Material