Author + information
- Received July 1, 2005
- Revision received September 8, 2005
- Accepted September 13, 2005
- Published online February 21, 2006.
- David A. Garcia, MD⁎,2,⁎ (, )
- Susan Regan, PhD†,
- Mark Crowther, MD‡,3 and
- Elaine M. Hylek, MD, MPH§,1
- ↵⁎Reprint requests and correspondence:
Dr. David A. Garcia, 2211 Lomas Northeast, MSC10 5550, Albuquerque, New Mexico 87131.
Objectives Among warfarin-treated patients with international normalized ratio (INR) >5, we sought to determine the risk of major bleeding within 30 days.
Background For warfarin-treated patients, the risk of bleeding increases as the INR rises, particularly if the INR exceeds 4. The 30-day risk of hemorrhage among outpatients with excessively prolonged INR values is unknown.
Methods To assess anticoagulation care in the U.S., a cohort of 6,761 patients taking warfarin was prospectively assembled from 101 participating sites (43% were community-based cardiology practices). From this cohort, 1,104 patients were identified with a first episode of INR >5.
Results A total of 979 met eligibility criteria; complete follow-up information was available for 976 (99.7%). Ninety-six percent (n = 937) of patients had an INR value between 5 and 9; 80% of INR values were <7. Thirteen patients (1.3%) experienced major hemorrhage during the 30-day follow-up period; among patients whose INR was >5 and <9, 0.96% experienced major hemorrhage. None of the bleeding events was fatal. Intervention with vitamin K was uncommon (8.7%). Warfarin doses were withheld for the majority of patients. Fifty percent of patients who were managed conservatively and retested on day 4 or 5 had an INR of 2.0 or less.
Conclusions For warfarin-treated outpatients presenting with an INR >5 and <9, the 30-day risk of major bleeding is low (0.96%). Intervention with vitamin K among asymptomatic patients presenting with an INR <9 is not routine practice in the U.S.
Warfarin is a commonly prescribed anticoagulant used to treat or prevent arterial and venous thrombosis. Because warfarin has a narrow therapeutic index, patients routinely undergo phlebotomy or capillary blood sampling to measure a prothrombin time, the laboratory assessment of anticoagulant effect. The prothrombin time is standardized between laboratories as the international normalized ratio (INR). For most indications, an INR value between 2.0 and 3.0 is targeted.
The narrow therapeutic index of warfarin is highlighted by the marked increase in the risk of major hemorrhage associated with INR values that exceed 4.0 (1–4). Although the association between an INR >4 and increased risk of hemorrhage is well documented, published evidence documenting the absolute 30-day bleeding risk for an individual, asymptomatic patient who presents with a supratherapeutic INR is limited. This uncertainty fuels patient fear and physician anxiety. In one prospective study of 114 asymptomatic patients presenting with INR values >6.0, 5 experienced a major hemorrhage during 14 days of follow-up (4.4%, 95% confidence interval [CI] 1.4% to 9.9%) compared to none of 268 patients whose INR was in the target range (5). In a retrospective case series, 3 of 23 patients (12.5%) whose INR values exceeded 10.0 had clinically important bleeding (follow-up interval was not specified) (6). In contrast, two other retrospective studies (in which length of follow-up was not specified) have suggested that for asymptomatic patients whose INR value is >6.0, the risk of major hemorrhage is <1% (7,8). All of the aforementioned studies were limited by small size.
It is widely accepted that patients taking warfarin who present with evidence of active bleeding and INR prolongation should be given coagulation factor replacement (most often in the form of plasma products) and vitamin K (9). For asymptomatic patients with warfarin-associated coagulopathy, however, management strategies vary. Although low-dose oral vitamin K has been shown to return an elevated INR to the therapeutic range more rapidly than placebo (10,11), many providers choose simply to withhold warfarin from such patients, allowing the INR to decline spontaneously (12). This inconsistency of vitamin K use is likely explained by the uncertainty surrounding the short-term risk of bleeding faced by these patients.
To estimate the 30-day risk of hemorrhage among patients with an excessively prolonged INR, we collected data prospectively from a large observational cohort. By better defining the bleeding risk in this population, we hoped to provide guidance to physicians caring for patients with excessive prolongation of their INR. Our secondary aim was to describe the management practices (e.g., frequency of vitamin K use and time to follow-up INR testing) employed by the anticoagulation providers who enrolled patients in this study.
A prospective cohort was assembled in order to assess anticoagulation care in the U.S. (13). Registered users of CoumaCare software (Bristol-Myers Squibb [formerly DuPont Pharmaceuticals], Princeton, New Jersey) were invited to participate by letter and through a study website. CoumaCare is a software program that was freely available and used by many anticoagulation management services for clinical purposes such as patient tracking, data entry, and record keeping. The program does not make dosing or follow-up interval recommendations. A total of 174 individual site registrations were received by McKesson HBOC BioServices (Rockville, Maryland), the designated registry specialists for the study. Of these, 101 sites had the technologic capability and the review board approval necessary to participate. All sites had at least one dedicated provider managing warfarin, usually within the setting of a community-based, physician group practice. Patients were invited to participate in the registry either by letter or in person (at the time of a routine appointment). To be eligible for inclusion in the registry, patients had to be 18 years of age or older and provide written informed consent. The participation of sites and patients was voluntary. For all centers, McKesson HBOC provided individual on-site training about how to recruit patients, obtain consent, and transmit data. Adverse event reporting was mandatory, and study personnel were trained to carry out such reporting with rigor sufficient to meet federal regulatory requirements. Enrollment began in April 2000 and ended in February 2002.
Encrypted data from each site were transmitted to the independent data-coordinating center weekly (McKesson HBOC BioServices). Missing data fields and data entry errors were flagged and resolved directly with the sites before data were transferred to the study investigators. A direct query from the data-coordinating center to the reporting site was triggered by any interval in INR testing that exceeded 45 days orany INR value >10 or <0.8. Resolution of the flag relating to the INR testing interval required validation of warfarin status and confirmation that the gap was not related to an adverse event. Study investigators were blinded to the identification and location of participating practices and patients.
Patients and outcomes
Patients whose INR measurement was ≥5.0 were identified. For patients who had more than one such qualifying INR value, only the first occurrence was considered. To be eligible, the index INR value had to: 1) be accompanied by a progress note (to classify provider management); and 2) occur more than 60 days before the site’s final study data transmission (to enable complete follow-up by McKesson HBOC).
For each eligible patient, an investigator (D.A.G. or E.M.H.) reviewed the anticoagulation progress notes and follow-up INR values recorded during the 60 days after the index INR measurement. The primary outcome of interest was major hemorrhage, defined as bleeding that was fatal, led to hospitalization with transfusion of at least 2 U of packed red blood cells, or occurred at a critical site (e.g., intracranial, retroperitoneal). All other bleeding was considered minor. Occurrence of an arterial or venous thromboembolic event in the 30-day follow-up period was also recorded. All major events were validated by an investigator (D.A.G. or E.M.H.) and directly verified with the site director by McKesson HBOC. For each patient, the following data were also recorded: number of days warfarin was withheld, number of days until the first repeat INR was performed, number of days until an INR value <4.0 was documented, and the presence or absence of documented vitamin K use.
The effect of vitamin K on the risk of major hemorrhage was assessed with logistic regression using a general estimating equation model to account for variation within sites. The model included terms for index INR, age, and vitamin K (14).
The study protocol was approved by Western Institutional Review Board (WIRB), Olympia, Washington, the institutional review board at Massachusetts General Hospital, and local review boards where they existed.
Role of the funding source
The funding source had no role in the collection, analysis, or interpretation of the data or in the decision to submit the study for publication.
In the anticoagulation management study, 6,761 patients were enrolled from 38 states providing 5,961 person-years of observation. Of the 101 clinical sites that participated, 98 were community-based physician office practices, and 3 were academic practices. Forty-three percent of the community-based sites were cardiology group practices. The most common indications for warfarin therapy included atrial fibrillation (AF) (52%), prosthetic heart valve (15%), and venous thromboembolism (14%). Overall, 83% of patients were 60 years of age or greater; 22% were age 80 years or greater. Among patients with AF, 46% had hypertension, 26% coronary artery disease, 15% diabetes mellitus, and 8% history of stroke.
From the overall cohort, 1,104 patients with a first observed INR ≥5 were identified; 979 (89%) met the eligibility criteria. Nearly all of the excluded patients were deemed ineligible because they had an elevated INR recorded fewer than 60 days before the site’s final data transmission. Of the 979 included patients, 39% were receiving warfarin for AF and 29% had a prosthetic heart valve. The mean age was 69 years (range 20 to 94), and 50% were women. Sixty-two patients (6%) had been taking warfarin for <3 months, 96% (n = 937) of the patients had an index INR value between 5 and 9, 80% of index INR values were <7 (Table 1).
Adverse events and documented vitamin K use
Of the 979 patients, follow-up was complete for 976 (99.7%). Overall, 13 patients (1.3% [95% CI 0.6 to 2.1]) experienced a major hemorrhage during the first 30 days after the index INR; the majority occurred within one week (Table 2).Most of these events were gastrointestinal in origin, and none was fatal. Among the 934 patients with an INR <9, 9 (0.96%) sustained a major hemorrhage. Four (9.5%) of the 42 patients whose INR was 9.0 or greater experienced a major bleed. Thirty-four patients (3.5%) reported minor bleeding; all such cases were self-limited except two patients whose epistaxis required cautery in the emergency department. One patient with a very elevated INR (but no sign of bleeding) was admitted to the hospital for observation and warfarin reversal. Thromboembolic events were infrequent. Three patients (0.3%) sustained an ischemic event during the 30-day follow-up period (two strokes and one peripheral arterial embolism).
Most patients were managed by withholding subsequent warfarin doses. For different subgroups, the mean number of days without warfarin is shown in Table 3.Vitamin K use, as documented in the progress notes in the Coumacare database, was not routine. The progress notes of 85 patients (8.7%) contained evidence that vitamin K was given. The dose of vitamin K was not available for five patients. Compared to patients with an index INR between 5 and 9, vitamin K use was more common among patients whose index INR was >9 (62% vs. 7%). When given, vitamin K was prescribed orally more than 90% of the time; the most common doses were 2.5 mg (60%), 5 mg (24%), and 1.25 mg (11%). In the U.S., vitamin K for oral administration is available only in 5-mg tablets. Of the 85 patients who were treated with vitamin K, one patient who received a 10-mg dose experienced an arterial embolus requiring embolectomy.
The effect of vitamin K on the risk of major hemorrhage was not statistically significant in the univariate analysis (odds ratio 0.85 [95% CI 0.11 to 6.67]; p = 0.88) or in the logistic model adjusting for index INR and age (odds ratio 0.76 [95% CI 0.10 to 6.00]; p = 0.80). The small number of major bleeding events in our study precludes definitive assessment of the relationship between vitamin K use and the risk of hemorrhage.
Patient follow-up and subsequent INR measurement
Of the 979 patients with an INR of 5.0 or greater, 12 patients had no subsequent INR measurements, 6 patients were taken off warfarin, 3 died of causes unrelated to warfarin, 2 transferred their warfarin management to another provider, and 1 patient was lost to follow-up. The time elapsed before the first repeat INR varied: 39% of patients were retested within three days, and 75% were retested within the first week. Of the 239 patients whose first follow-up INR occurred 8 or more days after the index INR, 120 (50%) were scheduled for follow-up >7 days after the index INR, 50 (21%) missed an earlier scheduled appointment, 50 (21%) had no documented reason for the follow-up interval chosen, 14 (6%) were hospitalized, and 5 (2%) had other reasons provided. For the patients whose warfarin was held (but for whom no vitamin K was prescribed), Figure 1shows the median INR result among patients tested on each of the first seven days after the index measurement. Only the first follow-up INR value for any patient was used for this analysis. Of the patients retested on day 4 or 5, nearly 50% had an INR of 2.0 or less.
The principal finding of our study is that for warfarin-treated asymptomatic outpatients with an INR value ≥5, the risk of major hemorrhage within 30 days is low (1.3%). It is noteworthy that these INR results were obtained in routine practice, and that 96% of index INR values were <9.0. The likelihood that our results fairly estimate the bleeding risk among such patients is high because our study included nearly 1,000 individuals identified from a cohort of 6,761 patients, 22% of whom were age 80 years or greater. Furthermore, our finding is consistent with two previously reported, smaller, single-center studies of 248 patients and 51 patients, respectively (7,8). The participation of 101 predominantly community-based anticoagulation sites in the present study enhances the generalizability of our findings.
The difference between the proportion of patients suffering major hemorrhage in our cohort and the higher proportion previously reported by Hylek et al. (5) may be explained by several important differences in the two studies. First, the previous study included only 114 patients, and the 95% CI surrounding the point estimate of major hemorrhage risk was wide. Second, the previous study included a population with characteristics that might be expected to increase the overall rate of hemorrhage: 100% of index INR values were >6 (14% were >10), the mean index INR was 8.1, the mean age was 71 years, and 13% of patients had been taking warfarin for <3 months. By comparison, the present study included a population in which only 3% of index INR values were >10, the mean index INR was 6.5, the mean age was 69, and only 6% of patients had been taking warfarin for <3 months. International normalized ratio, age, and the early phase of therapy have all been shown to be risk factors for major hemorrhage among patients taking warfarin (15–19); INR prolongation conveys the highest risk. During 14 days of follow-up in a previous study by Hylek et al. (5), there were no major hemorrhagic events among 268 patients whose INR was in the target range.
Administration of vitamin K, as documented in the CoumaCare database, was uncommon in our cohort despite evidence that low-dose oral vitamin K returns the INR to the normal range more quickly than placebo without causing “over-reversal” of warfarin’s anticoagulant effect (10,11,20–22). Published recommendations propose that low-dose oral vitamin K should be considered for patients with an INR >5 who are at risk for hemorrhage (23,24). However, the absence of high-quality evidence (coupled with the concern that vitamin K use may increase the risk of thromboembolism) likely explains the low frequency of vitamin K use among patients whose INR was <9.
We acknowledge that the patients at highest risk for bleeding may have avoided hemorrhagic events because they received vitamin K. However, because 894 individuals (91% of our cohort) were managed without vitamin K, the potential impact of any such selection bias on our results would be limited, particularly among patients whose INR was <9, for whom documented vitamin K use was particularly uncommon. In our analyses, vitamin K was not found to have a significant effect on the risk of major hemorrhage, but the small number of bleeding events precludes definitive assessment. Because participation in our study was voluntary and required written informed consent, the potential for selection bias does exist. However, our cohort is representative of community-based practice based on patient age and a prevalence of medical conditions that is very similar to that reported from other large ambulatory patient populations. To be eligible, patients had to be 18 years of age or older. Twenty-two percent of the 979 patients in our study cohort were age 80 years or greater. Patients with a prior history of bleeding would have been eligible, if they were still taking warfarin. The risk of major hemorrhage on warfarin is largely driven by anticoagulation intensity, and all patients in our study had an INR of 5.0 or greater. Because the CoumaCare software program does not incorporate decision aids or management tools, it is unlikely that use of this tracking system would have affected an individual’s risk of bleeding after an INR of ≥5.
In conclusion, this large cohort study indicates that for asymptomatic outpatients presenting with an INR between 5 and 9, the risk of major hemorrhage over the next 30 days is low (0.96%). The small number (n = 42) of patients whose INR was >9 precludes the extrapolation of our findings to this subgroup. Individuals presenting with INR values over 9 (and perhaps other subpopulations) may indeed have a higher 30-day risk of bleeding. Appropriate assessment of the potential benefit of selected (or routine) vitamin K administration will only be accomplished with randomized, placebo-controlled trials.
For a list of the practices and directors that participated in the study, please see the online version of this article.
↵1 Dr. Hylek has received research support from Bristol-Myers Squibb and Pfizer, and lecture honoraria from AstraZeneca.
↵2 Dr.Garcia has received consulting and lecture honoraria from AstraZeneca.
↵3 Dr. Crowther has received consulting honoraria from AstraZeneca.
This study was supported by a grant from Bristol-Myers Squibb (formerly DuPont Pharmaceuticals).
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- international normalized ratio
- Received July 1, 2005.
- Revision received September 8, 2005.
- Accepted September 13, 2005.
- American College of Cardiology Foundation