Oral anticoagulation with warfarin is widely used to treat venous and arterial thromboembolic disease (Ansell et al. Chest 133(6 suppl):160S-198S, 2008). Its administration is associated with a risk of intracranial hemorrhage (ICH), a devastating complication which usually results in death or severe disability (Fang et al. Am J Med,120(8):700-705, 2007; Rosand et al. Arch Intern Med, 164(8):880-884, 2004). The international normalized ratio (INR) is one of the factors which can help determine the risk of ICH in a given individual (Singer et al. Circ Cardiovasc Qual Outcomes, 2(4):297-304, 2009). Using the DoseResponse(®) patient database at our institution, we carried a retrospective nested matched case-control study to identify patient characteristics associated with the occurrence of ICH. The database was queried for the years 2004-2009. Each case was matched by month to four control patients having a routine INR determination for the monitoring of chronic anticoagulation. The following characteristics were captured: bleeding type, INR, age, sex, blood pressure, hemoglobin, creatinine, presence or history of pertinent comorbid conditions, intake of antiplatelet agents (aspirin or thienopyridine) and indication for anticoagulation. The relationship between those risk factors and the odds ratio of ICH was determined with conditional logistic regression. Fifty cases of ICH were retrieved. When correcting for pertinent variables, the odds ratio of ICH increased significantly for higher INR values, with a quadratic relationship noted in the model. Increasing mean blood pressure seemed to be associated with a higher risk of ICH, also in an exponential manner. Rising hemoglobin values on the other hand seemed to have a protective effect against such events. About 50% of cases of ICH occurred in or below the therapeutic INR range. The intake of antiplatelet agent was found to be associated with ICH in univariate analysis only. The INR is an important predictor for the incidence of ICH, but in this study an elevated measurement was found in only half of cases. Mean blood pressure appears to be another important determinant of the risk of ICH in the anticoagulated patient population.

1101 Background: oral anticoagulation with warfarin is used to treat venous and arterial thromboembolic disease. Its administration is associated with a risk of intracranial hemorrhage (ICH), a devastating complication which usually results in death or severe disability. The international normalized ratio (INR) is one of the factors which can help determine the risk of ICH in a given individual (Singer DE et al, Circ Cardiovasc Qual Outcomes 2009). Materials and methods: using the DoseResponse® patient database at our institution, we carried out a retrospective nested matched case-control study to identify patient characteristics associated with the occurrence of ICH. The database was queried for the years 2007 to 2009. Each case was matched by month to 4 control patients having a routine INR determination for the monitoring of chronic anticoagulation. The following characteristics were captured: INR, age, sex, systolic and diastolic blood pressure, hemoglobin, creatinine, history of pertinent medical conditions (hypertension, diabetes, heart failure, gastrointestinal bleeding, ischemic stroke, active cancer, substance abuse, cirrhosis), indication for anticoagulation (non-valvular atrial fibrillation, valvular atrial fibrillation, venous thromboembolism or other) and intake of antiplatelet agent. Blood pressure for cases was obtained from a medical encounter occurring before the bleeding event. The relationship between those risk factors and the odds ratio of ICH was determined with conditional logistic regression, using the SAS® 9.2 software platform. The initial approach consisted of stepwise regression with forward selection and backward elimination. Results: 31 cases of ICH were retrieved; they were matched to 124 controls. In the univariate analysis, the two groups differed significantly only in terms of their hemoglobin: 12.8 versus 13.5 g/dL for cases and controls, respectively (p=0.048). As for the INR, the mean value was 3.0 for cases vs 2.5 for controls. The distribution of this parameter was normal albeit more markedly skewed to the right for cases, with 3 values of 5.0 or more, compared to only one instance of this for controls. Most cases of ICH occurred in the setting of a therapeutic INR. The odds ratio (OR) of ICH (using the interval 2.01 to 2.50 as the reference) started increasing above an INR of 3.50, reaching its highest level in individuals with an INR value greater than 4.50 (OR=5.78, 95% CI=1.10-30.48). Mean blood pressures were similar between the two groups: 92 vs 89 mmHg for cases vs controls, respectively (p=0.252). The variables retained in the final regression model on the basis of statistical significance and clinical pertinence are shown in the table. The OR of ICH was 1.50 for increments of 1.0 in INR value (p=0.021), while it was 1.56 for increments of 10 mmHg in mean blood pressure (p=0.032). The presence of cancer, anemia and heart failure appeared to contribute to the risk of an event but the associations for those factors were not statistically significant. Conclusion: the INR is an important predictor for the incidence of ICH, but a supratherapeutic measurement is found only in a minority of cases; the risk of an event increases markedly with an INR above 3.5. Mean blood pressure is another important determinant of the risk of ICH in individuals on chronic warfarin therapy. Previous studies have shown that a diagnosis of hypertension is associated with an increased risk of ICH in the anticoagulated patient population (Berwaerts J et al, QJM 2000; Atrial Fibrillation Investigators, Arch Intern Med 1994; Singer DE et al, Ann Intern Med 2009), but to the knowledge of this author there has been no report describing the variation in this risk over the spectrum of mean blood pressures. This lends support to the generally accepted practice of aggressively treating arterial hypertension in patients on chronic oral anticoagulation. Multivariable Analysis Disclosures: No relevant conflicts of interest to declare.