Is the time in therapeutic range using the ratio of tests equivalent to the Rosendaal method?

Abstract

The percentage of time in therapeutic range (TTR) is a measure of anticoagulation quality with vitamin K antagonists (VKAs). The method most commonly used in clinical trials is the Rosendaal TTR. However, the application of this method in daily practice for clinical decision lacks appropriate instruments. We aimed to evaluate the percentage of tests within the target international normalized ratio (INR) (tests ratio) as a surrogate of Rosendaal TTR. We performed an observational and retrospective study to evaluate the TTR according to the Rosendaal method and tests ratio. We included all outpatients who attended the cardiology anticoagulation clinic of a Portuguese hospital (2011-2013), whose target INR was 2.0-3.0. Three hundred and seventy-seven VKA-treated patients followed for a mean 1.3 years were evaluated. Rosendaal methold and tests ratio significantly correlated (Rho Spearman 0.88, P < 0.001), but the Bland-Altman plot evaluation showed a clinically relevant data dispersion [95% confidence interval (95% CI) -12.9 to 23.1] around a mean difference in TTR -5.1% using the tests ratio method. The linear regression Passing-Bablok confirmed the existence of significant data dispersion and systematic differences. The tests ratio less than 60% had a sensitivity of 91.6%, specificity of 72.3%, positive predictive value (PPV) of 72.2% and negative predictive value (NPV) of 91.6%, for the diagnosis of patients inadequately anticoagulated (Rosendaal TTR <60%). Tests ratio had a c-statistics of 0.94 (95% CI 0.91-0.96). Number of tests in 6 months had a c-statistics of 0.70 (95% CI 0.65-0.75). Tests ratio underestimated TTR in 5% and was not considered equivalent to Rosendaal TTR due to the high variability between methods. Nevertheless, the use of tests ratio less than 60% may be a reasonable option to detect inadequate anticoagulation, as it is a sensitive method and excluded most of the patients with adequate control.