Anticoagulant management of atrial fibrillation: the influence of dosing algorithm and recall schedule on time in therapeutic range

Transcription

1 Family Practice, 2015, Vol. 32, No. 5, doi: /fampra/cmv066 Advance Access publication 5 August 2015 Health Service Research Anticoagulant management of atrial fibrillation: the influence of dosing algorithm and recall schedule on time in therapeutic range Y Swarna Nantha* Primary Care Research Initiative and Methods Education Department (PRIMED), Seremban Primary Care Clinic, Jalan Rasah, Seremban, Negeri Sembilan Darul Khusus, Malaysia. *Correspondence to Y Swarna Nantha, Non-Communicable Disease Clinic, Seremban Primary Care Clinic, Jalan Rasah, Seremban, Negeri Sembilan Darul Khusus, Malaysia; yogarabin@gmail.com Abstract Background. The quality of anticoagulation management in atrial fibrillation patients is reflected by the concept of time spent in therapeutic range (TTR). In a primary care setting, the implementation of a dose nomogram could help increase the mean TTR among these patients. Objective. This study compares the influence of a dose algorithm with an integrated recall on TTR prior to standard care and after the implementation of the protocol. Patients and methods. In a purposive sample of patients with AF, an uncontrolled before and after study design was utilized to measure the effects of the protocol on TTR. Demographic data, TTR levels, frequency of international normalized ratio (INR) within therapeutic range, clinician adherence to dose nomogram and warfarin dose changes were captured from consultations at the anticoagulation clinic. Results. A total of 152 patients with AF were entered into the final analysis. The increment in mean TTR in the after intervention phase (2.9%) was not statistically significant ( %, P = 0.252). The increase in the frequency of INR values within therapeutic range in the after intervention phase was significant ( %, P < 0.05) but with a very low effect size (r = 0.04). Conclusions. The implementation of a dose nomogram has the potential of reducing unnecessary dose changes for minor fluctuations in INR levels. The findings in this study needs to be confirmed in a future study involving other indications for anticoagulation, various regional primary care clinics and a larger population size. Key words. Atrial fibrillation, dose nomogram, predictors, primary care, TTR, warfarin. Introduction Clinicians from both tertiary and primary care practice believe that patients can benefit from warfarin management at a primary care setting (1). The efficacy of anticoagulant management at primary care centres has proven to be equal, if not superior, to similar treatment at a tertiary care centre (2 4). Consequently, new monitoring strategies were implemented at the primary care level and were found to be cost-effective in monitoring patients on anticoagulant treatment (4,5). Patients with atrial fibrillation (AF) require optimization of time spent in therapeutic range (TTR) while on vitamin K antagonists (VKA). TTR is a recognized way to measure the therapeutic effectiveness of VKA over time. TTR is measured by calculating the percentage of days spent in the therapeutic range (or percentage of time spent in the optimal international normalized ratio (INR) range for any indication for anticoagulation). The recent American College of Chest Physician guidelines have suggested the practice of three techniques in order to improve TTR namely (i) establishing anticoagulation clinics, (ii) application of computerized dose adjustment systems and (iii) patient self-monitoring (6). Majority of these methods warrant further evaluation before its implementation in a primary care setting could be finalized (5). The Author Published by Oxford University Press. All rights reserved. For permissions, please journals.permissions@oup.com. 514

2 Anticoagulant management of atrial fibrillation 515 Community-based practices had poorer performance in achieving target TTR when compared to treatment at specialized clinics (7). A recent study explored the relationship between adherence to an integrated dose protocol and recall schedule in the management of AF patient on VKAs (8). In the study, the algorithm suggested in previous research papers was modified to incorporate a structured maintenance recall rule (9,10). The analysis revealed a significant relationship between TTR and the integrated nomogram (dose protocol and recall schedule) when utilized at a primary care setting (8). This study attempts to evaluate the efficacy of an integrated decision nomogram in the management of anticoagulation maintenance in patients with AF at the primary care level (8). Hence, the study explores the relevance of a protocol designed to address anticoagulation monitoring and the efficiency of therapeutic decision making from a primary care perspective. Methods Setting The control of INR, defined by TTR, was examined in a randomly selected population in an urban primary care clinic. The sample comprises an adult population aged 18 and older treated for AF at the primary care clinic. The study was conducted at the outpatient department of the Seremban Primary Health Clinic, a regional primary care clinic in the state of Negeri Sembilan, Malaysia. The clinic is staffed by one family medicine specialist, two vocational trainees in family medicine and 30 medical officers. Study design A specialized anticoagulation clinic was established in the outpatient department to serve the needs of patients who require anticoagulation treatment. An uncontrolled before and after study design was utilized to determine the effects of intervention on TTR (11). The study consists of two stages. In the initial stage ( before phase), we determined the mean TTR of patients with AF (8). At this juncture, the warfarin dose was adjusted by doctors based on prior clinical experience, without the help of a decision support tool. During the second phase of the study ( after phase), the mean TTR was calculated after the introduction of a validated decision support tool in the form of a dosing and recall algorithm. The doctors were advised to adhere to the decision support tool when treating patients at the anticoagulation clinic. Inclusion criteria and exclusion criteria The study employed the purposive sampling of patients with AF receiving anticoagulation treatment at the outpatient department. Patients were excluded if the they were on other forms of anticoagulation besides warfarin, if they have been recently initiated on warfarin (less than 6 weeks) and if the monitored therapeutic INR range is not consistent with the suggested recommendations (INR range of 2 3 for AF patients) (9,10,12). Other exclusion criteria include patients who (i) had less than 6 months of anticoagulation treatment at the primary care clinic and (ii) had defaulted treatment/refuse to be a part of the study and (iii) information on INR levels were unretrievable/missing to aid statistical analysis. Time in therapeutic range TTR was utilized as the quality measurement of anticoagulation management in this study. TTR indicates how optimal the intensity of anticoagulation is maintained within the therapeutic range. The calculation of TTR was based on the modification of the Rosendaal linear interpolation method (13). The method assumes that there is a linear relationship between two consecutive INR results. This allows the quantification of the percentage of time for which the INR is below, within or above the optimal therapeutic range. Warfarin dose adjustment and recall period scheduling Doctors were instructed to adhere to the protocol presented in Tables A1 and A2 for dose/recall adjustments (9,14). A clinical course was conducted to ensure doctors in the anticoagulation clinic understood the function of the protocol. The strength of the dose of VKA was titrated according to the dose algorithm in the protocol (INR levels less than 2 is considered low and levels beyond 3 were considered high ). When irregularities in the INR readings were detected, patients were inquired about compliance to VKA, the consumption of vitamin K-rich foods, possible drug interactions (e.g. antibiotics) or the presence of any illnesses. The adherence to both dose algorithm and recall schedule was quantified by the investigator using a weighted pre-fixed scoring system. Data collection Data collection was performed in two phases. Firstly, data were collected when patients were monitored for a duration 6 months of standard care at the primary care clinic (the before intervention phase). In the subsequent 6 months, doctors were requested to adhere to the study protocol. The data during this phase of the study form the after intervention phase. In accordance to a previous study, the following time periods were excluded in our analysis, (i) the interval of more than 56 days between two consecutive INR readings and (ii) the period 1 week before and 3 weeks after maintenance interruption (9). This preserves data accuracy by (i) preventing the violation of the Rosendaal method employed in the calculation of TTR (13) and (ii) avoiding the inclusion of data that contributes to variation or instability in the INR readings due to VKA interruption. Prior to the intervention, recruited patients were followed up for at least 6 months at the primary care setting upon being discharged from the tertiary care hospital. This helped generate greater numbers of evaluable days/inr readings for each patient treated at the primary care clinic where the study was conducted. The concordance to the dosing algorithm was determined by quantifying proportions related to the adherence to the dose nomogram before and after the intervention took place. In the before category, adherence was calculated by calculating the similarity between usual practice and the recommendations defined by the dose algorithm. The proportions of decisions in agreement with the recall schedule were quantified in a similar manner both during the before and after phase. Data were then keyed into an automated anticoagulation monitoring spreadsheet in Microsoft Excel 2010 (version 14.0) by the investigator to obtain results for various parameters such as TTR, dose adherence, recall adherence and dose changes. Statistical analyses All statistical analysis was done using Statistical Package for Social Sciences (SPSS version 20; IBM Corp., USA). Continuous data are described as mean, median and standard deviation (SD). The obtained descriptive data were reported as mean and percentages. Paired t-test or Wilcoxon rank test was performed to compare the mean TTR, frequency of INR within range, consistency of dosing using the algorithm, proportions of adherence to the

3 516 Family Practice, 2015, Vol. 32, No. 5 recall schedule and percentage of dose changes before and after intervention. Multiple regression analysis was used to test whether the elements in the protocol could significantly predict changes in the TTR. All statistical analyses were done with 95% confidence intervals (CI) and the level of significance was set at P < Sample size of 149 patients was calculated based on the prevalence rate of AF patients on long-term anticoagulation at a primary care setting in Malaysia (73.8%) (8). A power analysis was done through the G*Power software (version 3.1.9, University of Kiel, Germany). The initial pilot study prior to intervention gave the effect size of 0.13 at a CI of 95% and the margin of error of 5% (8). This gave a statistical power of Results Demographic and descriptive details Out of the 253 patients who were invited, a total of 217 atrial fibrillation patients consented to participating in the study. However as shown in Tables 1 and 2, the final analysis involved ~152 of these patients (70.0%) who (i) had a complete data related to VKA therapy and (ii) had attended all appointments during the study. Thirty percent of the patients (n = 65) were excluded due to (i) total evaluable days less than 56 days and/or less than 3 INR readings to allow accurate statistical analysis (64%, n = 42), (ii) patients had less than 6 months of management at the primary care setting (18.5%, n = 12), (iii) patients were lost to follow up (10.7%, n = 7) and (iv) patients were on other anticoagulation medication other than VKA (3.1%, n = 2). Baseline characteristics of the cases in the study are shown in Table 3. The mean age of patients was 67.4 years (SD = 9.8). More than half of the patients were males (62.5%) while the rest were females (37.5%). Majority of the patients had hypertension (48.7%) followed by diabetes mellitus (28.3%), ischaemic heart disease 12.5%) and dyslipidaemia (11.2%). Table 1. Study population baseline characteristics (age, gender and concomitant illnesses) Characteristics of subject Values (SD) Age Mean age (SD), years 67.4 (9.8) Gender (n, %) Female 57 (37.5) Male 95 (62.5) Concomitant illnesses other than AF (n, %) Diabetes mellitus 43 (28.3) Hypertension 74 (48.7) Heart failure 7 (4.6) Cerebrovascular accident 14 (9.2) Hypothyroidism 1 (0.7) Hyperthyroidism 5 (3.3) Benign prostatic hyperplasia 5 (3.3) Chronic gouty arthritis 5 (3.3) Cardiomyopathy 2 (1.3) Ischaemic heart disease 19 (12.5) Bronchial asthma 2 (1.3) Dyslipidaemia 17 (11.2) Chronic obstructive airway disease 3 (2.0) Chronic rheumatic heart disease 4 (2.6) Chronic kidney disease 3 (2.0) End-stage renal failure 1 (0.7) Prostatic cancer 1 (0.7) Changes in TTR and frequency of INR within range Although there was an increase in overall mean TTR, a paired-sample t-test did not reveal a statistically significant increment in the before and after phase. There was a statistically significant increase in the frequency of INR values within therapeutic range following intervention with the dose nomogram before intervention (M = 50.0, SD = 20.6) and after intervention (M = 56, SD = 26.0, P < 0.05). However, the eta squared statistic (0.04) revealed a small effect size for this result. Adherence to dose protocol and recall schedule There was a statistically significant increase in both adherence to dose protocol and recall schedule among primary care physicians after the implementation of the anticoagulation clinic. A Wilcoxon sign rank test proved a significant increase in the median adherence to both dose algorithm (z = 6.647, P < 0.01, Md = %) and recall schedule (z = 9.740, P < 0.01, Md = %). The effect size for both predictors were at 0.38 and 0.56, respectively. Table 2. Results before and after the implementation of the study protocol in relation to TTR, adherence to protocol, variation in therapeutic range and changes in VKA dosing Characteristics of the subjects Mean (SD) or median (IQR) P value Effect size TTR (%) Mean TTR before intervention 57.5 (20.4) Mean TTR after intervention 60.4 (24.6) Adherence to dose algorithm (%) Median before intervention 75.0 (23.6) < Median after intervention 100 (20.0) Adherence to recall schedule (%) Median before intervention 37.5 (30.0) < Median after intervention 83.3 (33.3) Frequency of reading in therapeutic range (%) Mean before intervention 50.0 (20.6) < Mean after intervention 56.0 (26.0) Frequency of reading below therapeutic range (%) Median before intervention 42.9 (10.7) Median after intervention 37.5 (42.9) Frequency of reading above therapeutic range (%) Mean before intervention 21.8 (23.2) < Mean after intervention 14.3 (42.9) Frequency of change in VKA dose when above therapeutic range (%) Median before intervention < Reduce 50.0 No change 0.0 Median after intervention Reduce 0.0 No change 0.0 Frequency of change in VKA dose when below therapeutic range (%) Median before intervention < Increase 33.3 No change 50.0 Median after intervention Increase 0.0 No change 50.0 IQR, interquartile range.

4 Anticoagulant management of atrial fibrillation 517 Table 3. Regression analysis of TTR and frequency of INR values within therapeutic range in relation to dose algorithm Regression model Unstandardized equation ANOVA regression model Colinearity statistics Correlation statistics β adjusted Significance R 2 F Significance Tolerance VIF Pearson r Significance TTR Constant Dose algorithm < < <0.01 Frequency of INR within therapeutic range Constant Dose algorithm < < <0.01 VIF, variance inflation factor. Changes in the frequency of INR readings higher and lower than therapeutic range A statistically significant reduction in the frequency of INR values above therapeutic range was seen following intervention with the dose nomogram (z = 3.599, P < 0.01), with a medium effect size (r = 0.21). The median frequency of INR values higher than therapeutic range decreased from a pre-intervention level of 21.8% to a post-intervention level of 14.3%. There was no significant reduction in the frequency of INR readings below the therapeutic range following intervention. Changes in dosing of warfarin among doctors A Wilcoxon signed rank test revealed a statistically significant reduction in VKA dose changes when the INR readings were higher than the therapeutic range following intervention with the dose nomogram (z = 4.733, P < 0.01) with a medium effect size (r = 0.27). The median frequency of changes in VKA dose that occurred when there was an incidence of a high range INR values decreased from a pre-intervention level of 50% to a post-intervention level of 0%. There was a statistically significant reduction in VKA dose changes when the INR readings were lower than the therapeutic range following the intervention with the dose nomogram (z = 4.245, P < 0.01) with a medium effect size (r = 0.24). The median frequency of changes in VKA dose that occurred when there was an incidence of low range INR values decreased from a pre-intervention level of 33.3% to a post-intervention level of 0%. The effects of predictors on TTR and frequency of INR values within therapeutic range TTR was significantly correlated with the dose algorithm (r = 0.384, P < 0.01). However there was poor correlation with the recall schedule (r = 0.05, P = 0.58). The results of the regression show that the dose algorithm explained 14.7% of the variance. The dosing protocol significantly influenced the TTR (β = 0.69, P < 0.01). Hence, the regression equation derived from the following analysis was as follows: TTR = 0.69 (dose algorithm adherence) Frequency of INR values within therapeutic range was significantly correlated with the dose algorithm (r = 0.444, P < 0.01). Similarly, there was poor correlation with the recall schedule (r = 0.08, P = 0.076). Regression analysis found that the dose algorithm explained 14.7% of the variance. The dosing protocol significantly influenced the frequency of INR values within range (β = 0.84, P < 0.01). The regression equation derived from the following analysis was as follows: frequency of INR value within therapeutic range = 0.44 (dose algorithm adherence) Discussion Our study reveals that the reduction in the fluctuations of INR readings is possible through the establishment of an anticoagulation clinic and the implementation of a structured dose nomogram at a primary care setting. This was evidenced by lesser incidences of high and low INR readings in the after phase when compared to the before phase of the intervention. Among the two predictors that were tested in the regression analysis, only the dose algorithm had a statistically significant effect on TTR and frequency of INR readings within therapeutic range. The integration of a recall schedule to the dose algorithm seems to have very little impact on TTR and frequency of INR readings within therapeutic range. These results contradict an initial study that predicted the possibility of a minimal influence of a recall schedule on TTR if a prospective study was done utilizing a similar dose nomogram (8). Interestingly, the reluctance to fully comply to the recall schedule in the after phase (83.3%) could be attributed to the fact that the doctors felt that the decision to lengthen the interval between appointments could be detrimental to patients. This reluctance occurred in spite of being given specific instructions to strictly adhere to the recall schedule. A strict adherence to the dose protocol resulted in the reduction of unnecessary changes in VKA dose when there is a single incident of dose below or higher than the anticoagulation range. Although this did not result in a statistically significant increase in TTR, the implementation of the dose protocol helped to achieve greater numbers of INR readings that were within range. Strengths and limitations This study represents the first analysis of TTR trends related to a dose nomogram implementation in a primary care setting in Malaysia. A previous analysis was attempted at a tertiary care centre in Malaysia but did not include the examination of TTR and frequency of INR within therapeutic range (15). Most patients in the study had not achieved the target of 90 evaluable in the before phase or 30 evaluable days in the after phase within the third month of our study (9,10). Hence, our study was pre-emptively prolonged to 6 months in the both before and after phase to (i) obtain greater numbers of patients with optimum evaluable for inclusion in the final analysis and (ii) avoid inaccuracies in TTR readings by acquiring larger numbers of INR readings that reflect the actual trend of TTR. Although changes in TTR were not statistically significant after the intervention of a dose nomogram, the mean TTR level was elevated (by 2.5%) to an optimal range of 58 65% as specified by American College of Chest Physician guidelines to prevent

5 518 Family Practice, 2015, Vol. 32, No. 5 antithrombotic episodes (6). More importantly, there was (i) a statistically significant improvement in the frequency of INR readings that were within therapeutic range, (ii) minimization of unnecessary dose changes and (iii) reduction in higher INR values post-intervention. These findings could support efforts to set up formal INR monitoring at the primary care level in the country. Such a step could ensure the proper monitoring of long-term patients on anticoagulation and hence palliate the phenomenon of ping pong effect in INR readings (16). There are few limitations of this study. The recruitment of AF patients into this study involved a single regional primary care clinic. Hence, these results might not be representative of all primary care clinics in Malaysia although all patients attending this regional clinic were from smaller district clinics that had no access to INR blood investigation. Secondly, in the classification of AF patients, a broad term of AF was utilized, which was inclusive of patients with paroxysmal and chronic AF. Comparison with existing literature Although the median adherence to the dose algorithm and recall schedule were at 100% and 83.3%, respectively, the findings in this study were not superior to the usual care. These findings are comparable to another study conducted at multiple primary care centres which found no significant increase in mean TTR levels (10). In that particular study, there was an increase in the mean TTR level ( %, P = 0.73). Similarly, our study did not produce a statistically significant increase in mean TTR levels before and after intervention ( %, P = 0.252). On the other hand, when the dose nomogram was incorporated as a standard practice at a tertiary care centre, there were improvements in mean TTR levels among AF patients on warfarin ( %, P < 0.01) (9). Despite having no significant effect on TTR levels, employing a watchful waiting method instead of dose reduction or increment for mild fluctuations in INR values led to a significant increase frequency of mean INR levels within therapeutic range (M = %, P < 0.05). These findings are consistent with the result of a study that supports the idea of maintaining the same dose for patients with mild fluctuations in INR levels (17). Implications and future directions The natural biological variations in an isolated INR reading make monitoring of anticoagulant therapy difficult. By identifying intraindividual variations in INR, customization of anticoagulation management can be designed for each patient. Population-based biological variation in response to anticoagulation treatment can be determined by first identifying the total variation in INR reading in therapeutically stable patients (18,19). The implementation of protocols, as described in this study, does not only help increase TTR among patients but also help achieve greater numbers of stable patients that could be taken as a part of intra-individual biological analysis of INR fluctuations. Conclusion The preliminary results from this study indicate that there are benefits from the implementation of a dose nomogram in improving the frequency of INR values within therapeutic range at a primary care setting. The integration of a recall schedule did not improve TTR levels or the frequency of INR within therapeutic range in AF patients on VKAs. Acknowledgements The author would like to thank Dr Goh Pik Pin for her infinite support and inspiration during the course of writing this article. The author would also like to thank the Director General of Health Malaysia for allowing this study to be conducted at the research site and permission to publish this paper. Declaration Funding: none. Ethical approval: the study protocol was approved by the Malaysian Medical Research Ethic Committee (ethics committee approval number: NMRR , clinical protocol ID: 18365) and was conducted at the outpatient department of the Seremban Primary Health Clinic. Conflict of interest: none. References 1. Rodgers H, Sudlow M, Dobson R, Kenny RA, Thomson RG. Warfarin anticoagulation in primary care: a regional survey of present practice and clinicians views. Br J Gen Pract 1997; 47: Pell JP, McIver B, Stuart P, Malone DN, Alcock J. Comparison of anticoagulant control among patients attending general practice and a hospital anticoagulant clinic. Br J Gen Pract 1993; 43: Armero Simarro JM, Romero Cebrián MA, Ayuso Raya MC et al. Variability in the control of oral anticoagulant therapy between primary care and hospital in Albacete (2009). Rev Esp Salud Publica 2011; 85: Fitzmaurice DA, Hobbs FD, Murray ET et al. Oral anticoagulation management in primary care with the use of computerized decision support and near-patient testing: a randomized, controlled trial. Arch Intern Med 2000; 160: Hobbs FD, Fitzmaurice DA. Where should oral anticoagulation monitoring take place? Br J Gen Pract 1997; 47: Ageno W, Gallus AS, Wittkowsky A et al. Oral anticoagulant therapy: antithrombotic therapy and prevention of thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012; 141 (suppl 2): e44s 88S. 7. van Walraven C, Jennings A, Oake N, Fergusson D, Forster AJ. Effect of study setting on anticoagulation control: a systematic review and metaregression. Chest 2006; 129: Swarna Nantha Y. Anticoagulation management of atrial fibrillation: the relationship of prevailing adherence to a dose protocol and recall schedule on TTR in primary care. Heal Serv Res Manag Epidemiol 2014; 1: Kim YK, Nieuwlaat R, Connolly SJ et al. Effect of a simple two-step warfarin dosing algorithm on anticoagulant control as measured by time in therapeutic range: a pilot study. J Thromb Haemost 2010; 8: Nieuwlaat R, Eikelboom JW, Schulman S et al. Cluster randomized controlled trial of a simple warfarin maintenance dosing algorithm versus. J Thromb Thrombolysis 2013; 37: Grimshaw J, Campbell M, Eccles M, Steen N. Experimental and quasiexperimental designs for evaluating guideline implementation strategies. Fam Pract 2000; 17 (suppl 1): S Holbrook A, Schulman S, Witt DM et al. Evidence-based management of anticoagulant therapy: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012; 141 (suppl 2): e152s 84S. 13. Rosendaal FR, Cannegieter SC, van der Meer FJ, Briët E. A method to determine the optimal intensity of oral anticoagulant therapy. Thromb Haemost 1993; 69: Khot A, Polmear A. Venous thromboembolism. In: Khot A, Polmear A (eds.). Practical General Practice, 6th edn. London, UK: Elsevier, 2011, pp Freestone B, Rajaratnam R, Hussain N, Lip GYH. Admissions with atrial fibrillation in a multiracial population in Kuala Lumpur, Malaysia. Int J Cardiol 2003; 91:

6 Anticoagulant management of atrial fibrillation Wan Y, Heneghan C, Perera R et al. Anticoagulation control and prediction of adverse events in patients with atrial fibrillation: a systematic review. Circ Cardiovasc Qual Outcomes 2008; 1: Banet GA, Waterman AD, Milligan PE, Gatchel SK, Gage BF. Warfarin dose reduction vs watchful waiting for mild elevations in the international normalized ratio. Chest 2003; 123: Lassen JF, Kjeldsen J, Antonsen S, Hyltoft Petersen P, Brandslund I. Interpretation of serial measurements of international normalized ratio for prothrombin times in monitoring oral anticoagulant therapy. Clin Chem 1995; 41 (8 Pt 1): van den Besselaar AM, Fogar P, Pengo V et al. Biological variation of INR in stable patients on long-term anticoagulation with warfarin. Thromb Res 2012; 130: Appendix Table A1. Study protocol (dosing algorithm) Goal of INR in AF ( ) INR Action <1.5 a Increase weekly dose by 15%; repeat INR determination in 7 14 days a If falling or low on two or more occasions, increase weekly dose by 10%; repeat INR determination in 7 14 days No change a Do not hold warfarin. If rising or high on two or more occasions, decrease weekly dose by 10%; repeat INR determination in 7 14 days Hold for 1 day. Decrease weekly dose by 10%; repeat INR determination in 7 14 days Hold warfarin. Consider oral vitamin K 2 4 mg b if at increased risk of bleeding. If INR still high 24 hours later, consider giving 1 2 mg of additional oral vitamin K b and restart at lower dose (decrease weekly dose by 15%) when INR is therapeutic. Check INR weekly until stable >9.0 Hold warfarin and give oral vitamin K mg b. Monitor more frequently and repeat vitamin K if necessary Serious bleeding regardless of INR Hold dose and give intravenous vitamin K 10 mg and fresh frozen plasma, recombinant factor VIIa or prothrombin complex concentrates, depending on the urgency of the situation Adapted from Kim et al. (9). a Clinical and professional judgment may allow variation in the application of the algorithm. b The preferred method of vitamin K administration in non-emergency situations is orally. Avoid subcutaneous or intramuscular administration. Table A2. Study protocol (recall schedule) Scenario Action One high INR Recall in 7 14 days Stop treatment for 1 3 days (maximum 1 week in prosthetic valve patients One low INR Recall in 7 14 days One therapeutic INR Recall in 4 weeks Two therapeutic INRs Recall in 6 weeks Three therapeutic INRs Recall in 8 weeks Four therapeutic INRs Recall in 10 weeks Five therapeutic INRs Recall in 12 weeks Adapted from Khot and Polmear (14).