RESUMEN
OBJECTIVE: Information about rivaroxaban plasma level (RivLev) may guide treatment decisions in patients with acute ischemic stroke (AIS) and intracerebral hemorrhage (ICH) taking rivaroxaban. METHODS: In a multicenter registry-based study (Novel Oral Anticoagulants in Stroke Patients collaboration; ClinicalTrials.gov: NCT02353585) of patients with stroke while taking rivaroxaban, we compared RivLev in patients with AIS and ICH. We determined how many AIS patients had RivLev ≤ 100ng/ml, indicating possible eligibility for thrombolysis, and how many ICH patients had RivLev ≥ 75ng/ml, making them possibly eligible for the use of specific reversal agents. We explored factors associated with RivLev (Spearman correlation, regression models) and studied the sensitivity and specificity of international normalized ratio (INR) thresholds to substitute RivLev using cross tables and receiver operating characteristic curves. RESULTS: Among 241 patients (median age = 80 years, interquartile range [IQR] = 73-84; median time from onset to admission = 2 hours, IQR = 1-4.5 hours; median RivLev = 89ng/ml, IQR = 31-194), 190 had AIS and 51 had ICH. RivLev was similar in AIS patients (82ng/ml, IQR = 30-202) and ICH patients (102ng/ml, IQR = 51-165; p = 0.24). Trough RivLev(≤137ng/ml) occurred in 126/190 (66.3%) AIS and 34/51 (66.7%) ICH patients. Among AIS patients, 108/190 (56.8%) had RivLev ≤ 100ng/ml. In ICH patients, 33/51 (64.7%) had RivLev ≥ 75ng/ml. RivLev was associated with rivaroxaban dosage, and inversely with renal function and time since last intake (each p < 0.05). INR ≤ 1.0 had a specificity of 98.9% and a sensitivity of 25.7% to predict RivLev ≤ 100ng/ml. INR ≥ 1.4 had a sensitivity of 59.3% and specificity of 90.1% to predict RivLev ≥ 75ng/ml. INTERPRETATION: RivLev did not differ between patients with AIS and ICH. Half of the patients with AIS under rivaroxaban had a RivLev low enough to consider thrombolysis. In ICH patients, two-thirds had a RivLev high enough to meet the eligibility for the use of a specific reversal agent. INR thresholds perform poorly to inform treatment decisions in individual patients. Ann Neurol 2018;83:451-459.