Validation of an algorithm to predict decline in INR following warfarin cessation in patients undergoing invasive procedures

Patients on warfarin are required to withdraw from treatment for a fixed period (normally 5 days) prior to an invasive procedure. However, the anticoagulant effect of warfarin subsides at different rates among different patients, exposing some to increased risk of either thrombosis or bleeding. In a recent study in patients awaiting surgery, following warfarin cessation the INR declined slower over time in those with two CYP2C9 variant alleles, increasing age, weight and number of comorbidities and that INR decline was faster in those with higher maintenance INR value. Subsequently, we developed an algorithm which predicts INR decline in individual patients after 5 days of warfarin cessation. The current study validated the algorithm. An independent cohort of patients completing a short course of warfarin took part in the study. INR values for subsequent 9 days and CYP2C9 genotype were available. The predicted INR decline (INRday 1-INRday 5) was compared to the observed one (where an INR check on day 5 was unavailable, INR was estimated using a linear approximation model). There was a strong correlation between the decline in INR by day 5 and that predicted from the algorithm for the 117 patients (r = 0.949, p < 0.001). The algorithm was precise, with low degree of bias and variance of the prediction error. The algorithm can accurately predict the INR decline following warfarin cessation in individual adult patients. The use of this easily adoptable algorithm can reduce cancellation or delays of planned surgical procedures.