Review: anticoagulation with INRs less than 2 or between 3 and 5 increases risk for thromboembolic or hemorrhagic events.

Abstract

Question What is the effect of anticoagulation intensity on risk for thromboembolism or hemorrhage in patients taking oral anticoagulants (OAs)? Review scope Studies selected evaluated OA therapy for various indications and reported the number of outcome events and corresponding person-years of observation for 3 discrete ranges of international normalized ratios (INRs) (<2, 2 to 3, and >3). Patients taking both antiplatelet and anticoagulant medications were excluded. Outcomes were thromboembolism and hemorrhage; if severity of events was reported, those of minor severity were not included. Review methods MEDLINE, EMBASE/Excerpta Medica, CINAHL, and Cochrane Central Register of Controlled Trials (all to 2006), and Science Citation Index were searched for studies. 19 studies (n =80713, 56% men, mean follow-up 0.2 to 4.6 y) met the selection criteria: 10 retrospective cohort studies, 3 prospective cohort studies, and 6 randomized controlled trials (RCTs). 17 studies reported hemorrhagic events, including those that required hospitalization, blood transfusion, or surgery, or were intracranial or fatal; 16 reported thromboembolic events, including stroke, myocardial infarction, and systemic emboli. The meta-analysis compared outcomes by achieved INR ranges (reference INR range 2 to 3). Main results There were 98900 patient-years of follow-up across all trials. Meta-analysis showed that anticoagulation with INRs >3 increased risk for hemorrhage and anticoagulation with INRs <2 or >5 increased risk for thromboembolism (Table). When hemorrhage and thromboembolism were combined, anticoagulation with INRs <2, between 3 and 5, and >5 increased risk compared with the reference range (2 to 3) (Table). Significant heterogeneity between studies was detected. Conclusion In patients taking oral anticoagulants, anticoagulation with international normalized ratios <2 or between 3 and 5 increases risk for hemorrhage or thromboembolism compared with a reference range of 2 to 3. Anticoagulation with achieved international normalized ratios (INRs) in patients taking oral anticoagulants* Outcomes at mean 0.2 to 4.6 y follow-up Number of trials (n) Absolute risk/y (95% CI) INR <2 INR 2 to 3 INR 3 to 5 INR >5 Hemorrhage 17 (77913) 1.5% (0.7 to 3.0) 1.4% (0.9 to 2.3) 3.7% (2.2 to 6.3) 30% (15 to 61) Thromboembolism 16 (34706) 9% (6 to 13) 2.6% (1.8 to 3.6) 2.3% (1.5 to 3.4) 7% (3 to 14) Hemorrhage or thromboembolism 14 (31906) 11% (7 to 17) 4.3% (3.0 to 6.3) 7% (5 to 11) 52% (30 to 92) *Abbreviations defined in Glossary. Commentary The main conclusion of the meta-analysis by Oake and colleaguesthat optimal INR range is 2 to 3 across multiple indications for OA therapyis not controversial. The most recent authoritative guideline on OA use also suggests that dosing to an INR target of 2 to 3 is the safest and most effective range for most indications (1). The finding that rates of hemorrhage approximately double at an INR range of 3 to 5 and increase by as much as 10 to 20 times with INRs >5 is, unfortunately, borne out daily in clinical practice. It forms the basis of recommendations to administer small doses of oral vitamin K for asymptomatic INRs >5. The finding that rates of thromboembolic events may be increased at INRs >5 is controversial. There are several reasons why this may be an artifact. First, the review mixes studies with different designs, clinical indications, and target INR ranges. The observed heterogeneity suggests that combining these studies in meta-analysis might not be valid. Second, the results are heavily weighted by 2 large observational database studies where misclassification of OA exposure, INR result, and outcome is possible and comorbid conditions, drug interactions, and other confounders may be at play. Third, the number of events in this category is small, so the confidence intervals are wide. Could very high INRs be associated with clotting? It is plausible that a subgroup of patients with higher target INR ranges and rates of thromboembolism dominate this group (e.g., those with recurrent thromboembolism, cancer, older-generation mechanical valves, renal failure, or the antiphospholipid syndrome). With the additional worry that high INR results may be unreliablea median INR of 4 represented results as low as 2.2 or as high as 6.9 in different laboratories (2)there is ample incentive to diligently maintain OA therapy in the INR range of 2 to 3.