- Email: [email protected]
Single-dose adjustment versus no adjustment of warfarin in stably anticoagulated patients with an occasional international normalized ratio (INR) out of range
Thrombosis Research 125 (2010) 393–397
Contents lists available at ScienceDirect
Thrombosis Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t h r o m r e s
Regular Article
Single-dose adjustment versus no adjustment of warfarin in stably anticoagulated patients with an occasional international normalized ratio (INR) out of range S. Schulman ⁎, A. Melinyshyn, D. Ennis, L. Rudd-Scott Department of Medicine, McMaster University, Hamilton ON, Canada
a r t i c l e
i n f o
Article history: Received 18 March 2009 Received in revised form 1 July 2009 Accepted 7 July 2009 Available online 29 July 2009 Keywords: anticoagulants warfarin medication therapy management outpatients
a b s t r a c t Background: Well-controlled patients on warfarin may still have occasional International Normalized Ratios (INRs) outside the therapeutic range. It is unclear whether there is any benefit of a single-dose correction in this situation. Aim: To evaluate whether patients with very stable INR results should continue with the maintenance dose of warfarin without a single-dose correction, even when the result unexpectedly is moderately below or above the therapeutic range. Methods: A) We reviewed retrospectively 364 patients with unchanged maintenance dose for at least 6 months and an occasional INR outside the therapeutic range regarding decision on dosing and the effect on the next INR. B) We randomized 160 patients with at least 3 months of unchanged maintenance dose, an occasional INR deviating to a minimum of 1.5 or a maximum of 4.4 and unexplained or temporary, removable cause to a single-dose Change or No change. Follow-up INRs and telephone interviews were obtained after 2 weeks. Results: A) Retrospectively, the proportion of follow-up INRs outside the therapeutic range was 29.9% after No change, 27.1% after Increased dose and 25.7% after Skipped/reduced dose. However, the decision on No change was mainly taken in case of minimal INR deviations. B) Forty-eight (60%) of the patients randomized to Change were within the therapeutic range at follow-up versus 45 (56%) of those with No change, odds ratio 1.17 (95% confidence interval 0.59-2.30). For baseline INRs deviating down to 1.6 or up to 3.6 (therapeutic range, INR 2.0-3.0) the 2-week INRs did not differ between the groups. Conclusion: Our results suggest only a small or no difference between the two managements of an occasional INR out of range in terms of the 2-week follow-up INR. In stable patients on warfarin with an occasional INR outside the therapeutic range it seems reasonable to continue with the same dose without a single-dose change and perform a repeat test in about 2 weeks. (ClinicalTrials.gov number, NCT00814177.) © 2009 Elsevier Ltd. All rights reserved.
Introduction The management of patients on treatment with vitamin K antagonists (VKA) is a challenge due to the narrow therapeutic range and the difficulties to keep the prothrombin time within that range as a result of a multitude of interacting factors. Still, studies have shown that with excellent management and highly motivated patients the time in therapeutic range can be 75-92% [1]. A practical issue in daily clinical management of patients on VKA is whether an International Normalized Ratio (INR) outside the therapeutic range should result in a dose adjustment or only observation until the next INR is obtained. For the majority of patients a therapeutic range of International Normalized Ratio (INR) 2.0-3.0 is now recommended [1]. When a patient has had a series of INRs within the therapeutic range and Abbreviations: INR, International Normalized Ratio; VKA, vitamin K antagonists. ⁎ Corresponding author. Thrombosis Service, HHS-General Hospital, 237 Barton Street East, Hamilton ON, Canada L8L 2X2. Tel.: +1 905 5270271; fax: +1 905 5211551. E-mail address: [email protected] (S. Schulman). 0049-3848/$ – see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.thromres.2009.07.006
then without any identifiable explanation has a result outside the range there is often an intuitive wish to quickly correct this. Thus, in case of an INR above 3.0 the reaction would be to skip or reduce one dose of VKA and then continue as usual from the following day. Conversely, for an INR below 2.0 it may be tempting to give a single higher dose and then resume the current maintenance dose. Although this behaviour may bring the INR more rapidly back into the therapeutic range for the next few days, the effect on the following INR could be unimportant. The clinical benefit is probably negligible due to the very low risk of bleeding or thromboembolic complications from short periods outside the therapeutic range [2], although there are conflicting reports [3]. Some physicians advocate more frequent monitoring rather than inappropriate dose changes when the INR is for example within 0.5 INR from the therapeutic range [4–6]. Conversely, some manual or computerized algorithms suggest a change of the dose as soon as the INR is out of the desired range [7–10]. In a recent observational study Rose et al demonstrated that changes of the maintenance dose when the INR falls slightly outside the therapeutic range leads to a reduction of the time in therapeutic range [11].
394
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
We have in our practice observed another drawback of inappropriate single-dose changes. In spite of clear instructions that the suggested dose change applies only for a single day, many patients misunderstand and continue with this dose change every week on the same day. The true maintenance dose will therefore become different from what the clinic believes the patient is taking. In a cross-sectional review of our patients we found that at least 50% had no change to their maintenance dose during the previous 3 months and 33% had not had any change for at least 6 months [12]. Our hypothesis was that patients with very stable INR results should continue with the maintenance dose of warfarin without a single-dose correction, even when the result unexpectedly is moderately below or above the therapeutic range. We therefore performed two studies, a retrospective cohort (A) to describe the current situation and a prospective randomized trial (B) to test our hypothesis.
after the patient was randomized, using sealed envelopes with computer generated sequences. The nurse then instructed the patient accordingly to either continue with the usual dose a) without any changes or b) for elevated INR to skip/reduce one dose or for a low INR to take a larger dose (usually by 50%) the same day. The choice between skipping or reducing a dose was mainly practical. Thus, if the patient normally would take a 1/2 pill of whatever strength that day the instruction was to skip since splitting into quarters is difficult. In addition, for INRs of at least 4.0 skipping was preferred. Whenever reduced dose was ordered, it was always by 50%. A patient information and consent form was also sent by mail with a postage free return envelope. A laboratory visit for “follow-up INR” was scheduled after 2 weeks. This design was intended to mirror the typical routine at many anticoagulant clinics. At that time point the study nurse called the patient and asked about bleeding or thrombotic complications.
Methods
Statistical analysis
The studies were performed at the Anticoagulant Clinic of Thrombosis Service, HHS-General Hospital, which currently manages 1,400 patients, almost all on warfarin as the VKA agent. The patients have the prothrombin time analyzed at a laboratory that is most convenient for them and the INR results are faxed to our clinic, which informs the patients of the dose by telephone. During the study period there was no dosing software support and all INR results outside of the therapeutic range were assessed by one of three experienced physicians at the Anticoagulant Clinic. Patients were routinely instructed to have a repeat INR in 1-4 weeks, depending on the perceived risk of complications. Patients with INRs within the therapeutic range are usually monitored every 4 weeks.
The sample size for the randomized study was calculated on the following basis. From a retrospective review of 51 stable patients without any change in maintenance dose during the previous 6 months we knew that a) 13% of the INR results were below the range but above 1.4 and b) 17% of the results were above the range but below 4.5. Thus, 30% of the INR results during any period should make the patient eligible for inclusion. If the deviating INR was not caused by a long-standing effect, the “follow-up INR” would again have a 30% chance of falling outside of the therapeutic range. In order to detect a delta of 15% in out-of range INRs (i.e. an increase to 45%) with an alpha of 5% and power of 80% we would need 163 patients per group. We felt that a delta of 15% was acceptable, keeping in mind that the primary outcome was a surrogate endpoint. The study was planned for the sample size of 326 but due to competing studies, testing new dosing algorithms, the inclusion rate was much lower than expected. Point estimates for normal distributions are shown as mean and standard deviation and for skewed distributions with median and interquartile range. Comparisons of INRs were performed with MannWhitney test and for binomial data we used Pearson's Chi-square test with Yates correction. Possible associations between baseline characteristics and outcome were explored in univariate analysis. The studies were approved by the Hamilton Health Sciences/ Faculty of Health Sciences Research Ethics Board, the retrospective part as a chart review without need for consent.
A. Retrospective cohort In the retrospective study all dosing records for the 12-month period of July 2005-June 2006 were reviewed. Patients with unchanged maintenance dose and INR results within the therapeutic range for at least 6 months followed by an INR outside this range were then selected. We registered the most recent INR outside the range that was immediately preceded by such a period of stability (“baseline INR”), whether a dose was skipped or reduced (for high INR), an extra dose added (for low INR) or no action taken, and the following INR (qfollow-up INRq) result. Exclusion criteria were therapeutic range other than INR 2.0-3.0 or 2.5-3.5, patients with an INR below 1.5 or above 4.4, an interval exceeding 5 weeks to the “follow-up INR”. The proportion of qfollow-up INRsq within the therapeutic range was compared for patients with a single dose skipped/reduced/added versus no change of dose. B. Prospective randomized study The prospective study had a randomized, controlled, open-label design. Patients were eligible if they had at least 3 months of unchanged maintenance dose immediately prior to the qualifying, asymptomatic “baseline INR” outside the range, which again had to be INR 2.0-3.0 or 2.5-3.5. Exclusion criteria were age under 18; changes in diet, eating habits or concomitant medications that were likely to have a sustained effect (more than one week) on the prothrombin time, as captured in a telephone interview; INR results below 1.5 or above 4.4; or failure to obtain telephone consent. Approximately 15% of potentially eligible patients, according to the INRresults, were not available for screening by telephone and received their medication orders via their answering machine. The distribution of assessable patients is shown in Fig. 1. The study nurse called eligible patients and verified that no exclusion criteria were present. She then explained the study to the patient and obtained verbal consent by telephone. Immediately there-
Results A. Retrospective cohort In the retrospective study we reviewed 1321 records, of which 364 fulfilled the inclusion criteria, 157 with the therapeutic range of INR 2.0-3.0 and 107 with the more intensive range of INR 2.5-3.5. The deviations of baseline INR, when after a period of stability the result was outside the therapeutic range, are shown in Fig. 2. There were 215 baseline INRs below and 149 above the range. Overall, the decision was No change for 118 and Increase one dose for 97 below the range and likewise No change for 49 and Skip/reduce a dose for 100 above the range. The INRs resulting in No change were higher than those resulting in a single Increased dose for patients below the therapeutic range (mean INR 1.9 vs 1.875 for those with therapeutic range INR 2.03.0 and 2.2 vs 2.0 for those with a therapeutic range of 2.5-3.5; p b 0.001 for both). Similarly, the INRs that resulted in a Skipped/ reduced dose were higher than those resulting in No change for patients above the therapeutic range (mean INR 3.4 vs 3.2 for those with therapeutic range of 2.0-3.0 and 3.9 vs 3.6 for those with a therapeutic range of 2.5-3.5; p b 0.001 for both). The mean follow-up INRs for the same pairs of patient groups, after 28 ± 12 days, did not show any statistically significant differences. Overall, the proportion of
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
395
Fig. 1. The assessable population in the prospective randomized study (B). Patients were excluded due to long-term change in medication (11), long-term change in diet (3) or symptomatic INR out of range(1; epistaxis). Other reasons were inability to understand English (1) and confusion (1).
follow-up INRs outside of the therapeutic range was 29.9% after No change, 27.1% after Increased dose and 25.7% after Skipped/reduced dose (not significant). The decision was usually No change when for patients with therapeutic range INR 2.0-3.0 the result was 1.9 or 3.1 or for patients with therapeutic range 2.5-3.5 the INR was 2.2-2.4 or 3.6. For all INRs deviating more from the therapeutic range the decision was more often to perform a change. B. Prospective randomized study In the prospective study 160 patients were randomized during the period July 2006 – August 2008. All baseline characteristics are summarized in Table 1, and there were no statistically significant differences between the two groups. Increased intake of vegetables was paradoxically reported by two patients with baseline INR above the range, whereas the others with change in vegetable intake, missed dose or double dose of warfarin had the expected deviations. Antibiotics (amoxicillin – 3; levofloxacin – 1) were associated with decreased INR in two cases and increased in the other two. Other medications were reported with increased baseline INR in five of the nine cases. Two patients randomized to Change and 4 to No change had their repeat INR done already after one week in violation with the protocol. The deviations of baseline INR from the therapeutic range (Fig. 3) were less skewed within the subsets below and above the therapeutic range than in the retrospective study. The proportions of
INRs out of range (after 2 weeks) according to various subsets are shown in Table 2. Overall, among patients randomized to Change 35 (44%) were outside the therapeutic range at follow-up compared to 32 (40%) of those randomized to No change, odds ratio 1.17 (95% confidence interval 0.59-2.30), p = 0.75. When for every baseline INR the follow-up mean INRs were compared in the subset of patients with therapeutic range 2.0-3.0 there were minimal differences up to baseline INR 3.6, above which only 20 patients were recruited (Fig. 4). For the subset with therapeutic range INR 2.5-3.5 the number of patients at every INR was too small for meaningful comparisons. None of the recorded baseline characteristics were associated with the follow-up INR being within the range. In a post-hoc analysis, the median discrepancy of the baseline INR from the closest end of the therapeutic range was 0.3 and 0.4 for patients with follow-up INR within or outside the therapeutic range, respectively (p = 0.062). Adverse events registered during the 2-week observation period were two in the group randomized to No change (ventricular tachycardia triggering the implanted cardioverter/defibrillator; food poisoning) and one in the group randomized to Change (bleeding from penis, baseline INR 2.0, follow-up INR 3.0, therapeutic range 2.5-3.5). Discussion We have in a prospective, randomized study (B) demonstrated that for patients with a stable maintenance dose of warfarin, defined
Fig. 2. Distribution of the qualifying “baseline INRs”, shown as deviations from therapeutic range, in the retrospective cohort (A).
396
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
Table 1 Baseline characteristics for the patients in the prospective randomized study (B). Characteristic
No change of dose
Changed dose
N Male sex, n (%) Age⁎ Indication (main) Atrial fibrillation Mechanical valve, aortic Mechanical valve, mitral Venous thromboembolism Other INR range 2.0-3.0, n (%) Years on warfarin⁎ Months on same dose⁎ Maintenance dose, mg/week⁎ Previous bleeding event Reason for INR deviation Unknown Missed dose Double dose (mistake) Less vegetables in diet More vegetables in diet Antibiotic treatment Other medication change Other reason§ Baseline INR⁎ Below range, INR 2.0-3.0 Above range, INR 2.0-3.0 Below range, INR 2.5-3.5 Above range, INR 2.5-3.5
80 46 (57.5) 70.5 (64-78)
80 49 (61.3) 67.1 (57-79)
Table 2 Proportion of patients outside of therapeutic range after 2 weeks according to randomization (study B). Intensity
Deviation
N
Dosing
Out of range (%)
2.0-3.0
Below range
28 34 38 32 10 6 4 8
No change Increased No change Skipped⁎ No change Increased No change Skipped⁎
36 26 45 47 60 50 50 63
Above range 33 20 13‡ 10 4 66 (82.5) 4.5 (2-10) 7 (5-9) 30 (24.8-37.5) 1
32 24† 13‡ 5 6 66 (82.5) 5 (2-10.2) 7 (5-10.2) 31 (22.5-45.6) 6
43 5 2 12 9 2 6 1
53 3 1 7 9 2 3 2
1.7 (1.6-1.8) 3.4 (3.2-3.7) 1.8 (1.6-2.1) 3.9 (3.8-3.9)
1.7 (1.6-1.8) 3.4 (3.2-3.6) 1.8 (1.6-2.1) 4.0 (3.7-4.1)
None of the differences are statistically significant. ⁎Median and interquartile range. † Two patients had also atrial fibrillation. ‡One patient had also atrial fibrillation. §Other reasons were diarrhea, intake of alcohol or vitamins.
as 3 months without any change, and with an asymptomatic INR result slightly outside of the therapeutic range, there is no appreciable difference between changing a single dose or continuation without any change. In addition we showed in the retrospective study (A) that experienced physicians at an anticoagulation clinic only were prone to order watchful waiting if the INR result was 0.1 INR from the therapeutic range, except for patients with therapeutic range 2.5-3.5 and subtherapeutic INR, where a deviation down to 2.2 appeared acceptable. This resulted in confounding by severity [13] and illustrates the need for randomized, controlled studies. To our knowledge this is the first study where patients were randomized to different dosing in this context. Banet et al described a similar experience in 231 patients on warfarin, where their management had been randomized to anticoagulation service with recommendation to avoid changing the dose or to primary care providers without particular guidelines [14]. In their study only patients with
2.5-3.5
Below range Above range
p-value† 0.61 0.86 1.0 1.0
⁎Skipped or reduced dose. †Pearson's Chi-square with Yates correction, compared to the subset on the preceding line.
INR slightly above the range (INR 3.2, 3.3 or 3.4) were included and there was no uniform management in either group. However, they concluded that in case of an INR of 3.2 or 3.3 those managed at anticoagulation service were more likely to have a follow-up INR within the range than those managed by primary care providers. In our study we had wider margins for eligible excessive INRs but we also required a longer history of stability than in the study by Banet et al (3 months vs. 30 days) [14]. Furthermore, we tried to verify in the telephone interview that the patients did not have a new and continuing risk factor for altered warfarin effect, e.g. long-term medication with amiodarone, antibiotics or long-term change in diet. The proportion of follow-up INRs outside the therapeutic range was higher (40-44%) than the 30% anticipated in the sample size calculation. On the other hand the difference between the groups was minimal. With these proportions, an alpha of 5% and power of 80% we would have required about 2,400 patients per group to show that No change is worse than Change. Due to competing studies on dosing algorithms and on alternative monitoring intervals the recruitment to this study became slower than expected and a decision was therefore made to discontinue recruitment at 160 patients. One of the competing studies recruits patients with at least 6 months of unchanged dose of warfarin. It is possible that without this competition we would have seen a higher proportion of follow-up INRs within the therapeutic range. With the actual sample size of 160 in our study and the proportion of INRs out-of-range in the control group (40%) we ended up with a power of only 47.5% to demonstrate a delta of 15%. There is a power of 80% to demonstrate a delta of 22%, i.e. an increase of out-ofrange INRs from 40% to 62%. The selected time point for follow-up can be discussed. It is obvious that an added or skipped dose of warfarin will affect the INR during the next couple of days, but that is irrelevant in the perspective of this study. Patients with a small deviation of INR are not routinely re-tested after 2-
Fig. 3. Distribution of the qualifying “baseline INRs”, shown as deviations from therapeutic range, in the prospective randomized study (B).
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
397
modification of the maintenance dose is probably needed but should take into account the magnitude of the deviation. Unnecessary dose changes make the treatment more complex for the patients and should be avoided when no benefit can be demonstrated. When the unexpected INR deviation is more than 0.5 it may be advantageous to perform a change of the maintenance dose. Further studies on all these aspects of warfarin dosing should be encouraged. Conflict of interest statement None. Acknowledgement We are grateful for the assistance by Eric Danzig with the retrospective cohort. References
Fig. 4. Mean “follow-up INR” and standard deviation according to dosing allocation for each INR of “baseline INR” in the prospective randomized study (B). Symbols have been moved slightly sideways in opposite directions for the two randomization arms to avoid overlapping. INR categories with less than 3 data points are not presented.
3 days. In clinical practice the next test would be after 1-2 weeks and our results reflect what is seen then – that approximately 60% of those who were unexpectedly out of range are back in the therapeutic range after 2 weeks, irrespectively of what was done initially. Finally, we used a surrogate outcome, but a clinical event during this short observation period is very unlikely, as illustrated by our adverse events and would have required a completely impractical sample size. Studies on patients with chronic anticoagulation and brief periods of excessive or subtherapeutic INR have shown a low risk of clinically important events and no need for correction with vitamin K or bridging therapy, respectively [2,15]. Due to the limited number of patients in our study we are unable to rule out a 15% difference in the primary outcome and therefore the result is not definitive. However, a large effect can be excluded and therefore we conclude that in stable patients with an INR result up to 0.5 INR from the therapeutic range there is no or only a small difference in the followup INR at 2 weeks if a single dose is changed versus no change at all. This supports the concept of maintaining the same dose without a singledose change that may be confusing for the patient. A repeat INR should be performed after approximately 2 weeks, provided that the reason for the deviation is temporary and removed or that it is unexplained. If the repeat INR still is outside the range in the same direction a small
[1] Ansell J, Hirsh J, Hylek E, Jacobson A, Crowther M, Palareti G. Pharmacology and management of the vitamin K antagonists: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Chest 2008;133: 160S–98S. [2] Crowther MA, Ageno W, Garcia D, Wang L, Witt DM, Clark NP, et al. Oral vitamin K vs. placebo in the correction of excessively anticoagulated patients on warfarin: results of a randomized, controlled trial. Ann Intern Med 2009;150:293–300. [3] Hylek EM, Chang YC, Skates SJ, Hughes RA, Singer DE. Prospective study of the outcomes of ambulatory patients with excessive warfarin anticoagulation. Arch Intern Med 2000;160:1612–7. [4] Gunneman T, Ruybalid RL, Jacobson AK, Snell B, Zucker ML, LaDuca FM, et al. Frequent prothrombin time testing reduces inappropriate warfarin dose changes. Thromb Haemost 1999;82:676 (suppl) (abstract). [5] Gage BF, Fihn SD, White RH. Management and dosing of warfarin therapy. Am J Med 2000;109:481–8. [6] Waterman AD, Milligan PE, Banet GA, Gatchel SK, Gage BF. Establishing and running an effective telephone-based anticoagulation service. J Vasc Nurs 2001;19: 126–32. [7] Britt RP, James AH, Raskino CL, Thompson SG. Factors affecting the precision of warfarin treatment. J Clin Pathol 1992;45:1003–6. [8] Dalere GM, Coleman RW, Lum BL. A graphic nomogram for warfarin dosage adjustment. Pharmacotherapy 1999;19:461–7. [9] Horton JD, Bushwick BM. Warfarin therapy: evolving strategies in anticoagulation. Am Fam Phys 1999;59:635–46. [10] Ryan PJ, Gilbert M, Rose PE. Computer control of anticoagulant dose for therapeutic management. Bmj 1989;299:1207–9. [11] Rose AJ, Ozonoff A, Berlowitz DR, Henault LE, Hylek EM. Warfarin dose management affects INR control. J Thromb Haemost 2009;7:94–101. [12] Schulman S. A preliminary study on the feasibility of extending the interval between prothrombin time tests for patients on warfarin therapy. Haematologica 2005;90:132 (abstr 0336). [13] Salas M, Hofman A, Stricker BHC. Confounding by indication: an example of variation in the use of epidemiologic terminology. Am J Epidemiol 1999;149. [14] 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;123:499–503. [15] Clark NP, Witt DM, Delate T, Trapp M, Garcia D, Ageno W, et al. Thromboembolic consequences of subtherapeutic anticoagulation in patients stabilized on warfarin therapy: the low INR study. Pharmacother 2008;8:960–7.
Contents lists available at ScienceDirect
Thrombosis Research j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t h r o m r e s
Regular Article
Single-dose adjustment versus no adjustment of warfarin in stably anticoagulated patients with an occasional international normalized ratio (INR) out of range S. Schulman ⁎, A. Melinyshyn, D. Ennis, L. Rudd-Scott Department of Medicine, McMaster University, Hamilton ON, Canada
a r t i c l e
i n f o
Article history: Received 18 March 2009 Received in revised form 1 July 2009 Accepted 7 July 2009 Available online 29 July 2009 Keywords: anticoagulants warfarin medication therapy management outpatients
a b s t r a c t Background: Well-controlled patients on warfarin may still have occasional International Normalized Ratios (INRs) outside the therapeutic range. It is unclear whether there is any benefit of a single-dose correction in this situation. Aim: To evaluate whether patients with very stable INR results should continue with the maintenance dose of warfarin without a single-dose correction, even when the result unexpectedly is moderately below or above the therapeutic range. Methods: A) We reviewed retrospectively 364 patients with unchanged maintenance dose for at least 6 months and an occasional INR outside the therapeutic range regarding decision on dosing and the effect on the next INR. B) We randomized 160 patients with at least 3 months of unchanged maintenance dose, an occasional INR deviating to a minimum of 1.5 or a maximum of 4.4 and unexplained or temporary, removable cause to a single-dose Change or No change. Follow-up INRs and telephone interviews were obtained after 2 weeks. Results: A) Retrospectively, the proportion of follow-up INRs outside the therapeutic range was 29.9% after No change, 27.1% after Increased dose and 25.7% after Skipped/reduced dose. However, the decision on No change was mainly taken in case of minimal INR deviations. B) Forty-eight (60%) of the patients randomized to Change were within the therapeutic range at follow-up versus 45 (56%) of those with No change, odds ratio 1.17 (95% confidence interval 0.59-2.30). For baseline INRs deviating down to 1.6 or up to 3.6 (therapeutic range, INR 2.0-3.0) the 2-week INRs did not differ between the groups. Conclusion: Our results suggest only a small or no difference between the two managements of an occasional INR out of range in terms of the 2-week follow-up INR. In stable patients on warfarin with an occasional INR outside the therapeutic range it seems reasonable to continue with the same dose without a single-dose change and perform a repeat test in about 2 weeks. (ClinicalTrials.gov number, NCT00814177.) © 2009 Elsevier Ltd. All rights reserved.
Introduction The management of patients on treatment with vitamin K antagonists (VKA) is a challenge due to the narrow therapeutic range and the difficulties to keep the prothrombin time within that range as a result of a multitude of interacting factors. Still, studies have shown that with excellent management and highly motivated patients the time in therapeutic range can be 75-92% [1]. A practical issue in daily clinical management of patients on VKA is whether an International Normalized Ratio (INR) outside the therapeutic range should result in a dose adjustment or only observation until the next INR is obtained. For the majority of patients a therapeutic range of International Normalized Ratio (INR) 2.0-3.0 is now recommended [1]. When a patient has had a series of INRs within the therapeutic range and Abbreviations: INR, International Normalized Ratio; VKA, vitamin K antagonists. ⁎ Corresponding author. Thrombosis Service, HHS-General Hospital, 237 Barton Street East, Hamilton ON, Canada L8L 2X2. Tel.: +1 905 5270271; fax: +1 905 5211551. E-mail address: [email protected] (S. Schulman). 0049-3848/$ – see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.thromres.2009.07.006
then without any identifiable explanation has a result outside the range there is often an intuitive wish to quickly correct this. Thus, in case of an INR above 3.0 the reaction would be to skip or reduce one dose of VKA and then continue as usual from the following day. Conversely, for an INR below 2.0 it may be tempting to give a single higher dose and then resume the current maintenance dose. Although this behaviour may bring the INR more rapidly back into the therapeutic range for the next few days, the effect on the following INR could be unimportant. The clinical benefit is probably negligible due to the very low risk of bleeding or thromboembolic complications from short periods outside the therapeutic range [2], although there are conflicting reports [3]. Some physicians advocate more frequent monitoring rather than inappropriate dose changes when the INR is for example within 0.5 INR from the therapeutic range [4–6]. Conversely, some manual or computerized algorithms suggest a change of the dose as soon as the INR is out of the desired range [7–10]. In a recent observational study Rose et al demonstrated that changes of the maintenance dose when the INR falls slightly outside the therapeutic range leads to a reduction of the time in therapeutic range [11].
394
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
We have in our practice observed another drawback of inappropriate single-dose changes. In spite of clear instructions that the suggested dose change applies only for a single day, many patients misunderstand and continue with this dose change every week on the same day. The true maintenance dose will therefore become different from what the clinic believes the patient is taking. In a cross-sectional review of our patients we found that at least 50% had no change to their maintenance dose during the previous 3 months and 33% had not had any change for at least 6 months [12]. Our hypothesis was that patients with very stable INR results should continue with the maintenance dose of warfarin without a single-dose correction, even when the result unexpectedly is moderately below or above the therapeutic range. We therefore performed two studies, a retrospective cohort (A) to describe the current situation and a prospective randomized trial (B) to test our hypothesis.
after the patient was randomized, using sealed envelopes with computer generated sequences. The nurse then instructed the patient accordingly to either continue with the usual dose a) without any changes or b) for elevated INR to skip/reduce one dose or for a low INR to take a larger dose (usually by 50%) the same day. The choice between skipping or reducing a dose was mainly practical. Thus, if the patient normally would take a 1/2 pill of whatever strength that day the instruction was to skip since splitting into quarters is difficult. In addition, for INRs of at least 4.0 skipping was preferred. Whenever reduced dose was ordered, it was always by 50%. A patient information and consent form was also sent by mail with a postage free return envelope. A laboratory visit for “follow-up INR” was scheduled after 2 weeks. This design was intended to mirror the typical routine at many anticoagulant clinics. At that time point the study nurse called the patient and asked about bleeding or thrombotic complications.
Methods
Statistical analysis
The studies were performed at the Anticoagulant Clinic of Thrombosis Service, HHS-General Hospital, which currently manages 1,400 patients, almost all on warfarin as the VKA agent. The patients have the prothrombin time analyzed at a laboratory that is most convenient for them and the INR results are faxed to our clinic, which informs the patients of the dose by telephone. During the study period there was no dosing software support and all INR results outside of the therapeutic range were assessed by one of three experienced physicians at the Anticoagulant Clinic. Patients were routinely instructed to have a repeat INR in 1-4 weeks, depending on the perceived risk of complications. Patients with INRs within the therapeutic range are usually monitored every 4 weeks.
The sample size for the randomized study was calculated on the following basis. From a retrospective review of 51 stable patients without any change in maintenance dose during the previous 6 months we knew that a) 13% of the INR results were below the range but above 1.4 and b) 17% of the results were above the range but below 4.5. Thus, 30% of the INR results during any period should make the patient eligible for inclusion. If the deviating INR was not caused by a long-standing effect, the “follow-up INR” would again have a 30% chance of falling outside of the therapeutic range. In order to detect a delta of 15% in out-of range INRs (i.e. an increase to 45%) with an alpha of 5% and power of 80% we would need 163 patients per group. We felt that a delta of 15% was acceptable, keeping in mind that the primary outcome was a surrogate endpoint. The study was planned for the sample size of 326 but due to competing studies, testing new dosing algorithms, the inclusion rate was much lower than expected. Point estimates for normal distributions are shown as mean and standard deviation and for skewed distributions with median and interquartile range. Comparisons of INRs were performed with MannWhitney test and for binomial data we used Pearson's Chi-square test with Yates correction. Possible associations between baseline characteristics and outcome were explored in univariate analysis. The studies were approved by the Hamilton Health Sciences/ Faculty of Health Sciences Research Ethics Board, the retrospective part as a chart review without need for consent.
A. Retrospective cohort In the retrospective study all dosing records for the 12-month period of July 2005-June 2006 were reviewed. Patients with unchanged maintenance dose and INR results within the therapeutic range for at least 6 months followed by an INR outside this range were then selected. We registered the most recent INR outside the range that was immediately preceded by such a period of stability (“baseline INR”), whether a dose was skipped or reduced (for high INR), an extra dose added (for low INR) or no action taken, and the following INR (qfollow-up INRq) result. Exclusion criteria were therapeutic range other than INR 2.0-3.0 or 2.5-3.5, patients with an INR below 1.5 or above 4.4, an interval exceeding 5 weeks to the “follow-up INR”. The proportion of qfollow-up INRsq within the therapeutic range was compared for patients with a single dose skipped/reduced/added versus no change of dose. B. Prospective randomized study The prospective study had a randomized, controlled, open-label design. Patients were eligible if they had at least 3 months of unchanged maintenance dose immediately prior to the qualifying, asymptomatic “baseline INR” outside the range, which again had to be INR 2.0-3.0 or 2.5-3.5. Exclusion criteria were age under 18; changes in diet, eating habits or concomitant medications that were likely to have a sustained effect (more than one week) on the prothrombin time, as captured in a telephone interview; INR results below 1.5 or above 4.4; or failure to obtain telephone consent. Approximately 15% of potentially eligible patients, according to the INRresults, were not available for screening by telephone and received their medication orders via their answering machine. The distribution of assessable patients is shown in Fig. 1. The study nurse called eligible patients and verified that no exclusion criteria were present. She then explained the study to the patient and obtained verbal consent by telephone. Immediately there-
Results A. Retrospective cohort In the retrospective study we reviewed 1321 records, of which 364 fulfilled the inclusion criteria, 157 with the therapeutic range of INR 2.0-3.0 and 107 with the more intensive range of INR 2.5-3.5. The deviations of baseline INR, when after a period of stability the result was outside the therapeutic range, are shown in Fig. 2. There were 215 baseline INRs below and 149 above the range. Overall, the decision was No change for 118 and Increase one dose for 97 below the range and likewise No change for 49 and Skip/reduce a dose for 100 above the range. The INRs resulting in No change were higher than those resulting in a single Increased dose for patients below the therapeutic range (mean INR 1.9 vs 1.875 for those with therapeutic range INR 2.03.0 and 2.2 vs 2.0 for those with a therapeutic range of 2.5-3.5; p b 0.001 for both). Similarly, the INRs that resulted in a Skipped/ reduced dose were higher than those resulting in No change for patients above the therapeutic range (mean INR 3.4 vs 3.2 for those with therapeutic range of 2.0-3.0 and 3.9 vs 3.6 for those with a therapeutic range of 2.5-3.5; p b 0.001 for both). The mean follow-up INRs for the same pairs of patient groups, after 28 ± 12 days, did not show any statistically significant differences. Overall, the proportion of
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
395
Fig. 1. The assessable population in the prospective randomized study (B). Patients were excluded due to long-term change in medication (11), long-term change in diet (3) or symptomatic INR out of range(1; epistaxis). Other reasons were inability to understand English (1) and confusion (1).
follow-up INRs outside of the therapeutic range was 29.9% after No change, 27.1% after Increased dose and 25.7% after Skipped/reduced dose (not significant). The decision was usually No change when for patients with therapeutic range INR 2.0-3.0 the result was 1.9 or 3.1 or for patients with therapeutic range 2.5-3.5 the INR was 2.2-2.4 or 3.6. For all INRs deviating more from the therapeutic range the decision was more often to perform a change. B. Prospective randomized study In the prospective study 160 patients were randomized during the period July 2006 – August 2008. All baseline characteristics are summarized in Table 1, and there were no statistically significant differences between the two groups. Increased intake of vegetables was paradoxically reported by two patients with baseline INR above the range, whereas the others with change in vegetable intake, missed dose or double dose of warfarin had the expected deviations. Antibiotics (amoxicillin – 3; levofloxacin – 1) were associated with decreased INR in two cases and increased in the other two. Other medications were reported with increased baseline INR in five of the nine cases. Two patients randomized to Change and 4 to No change had their repeat INR done already after one week in violation with the protocol. The deviations of baseline INR from the therapeutic range (Fig. 3) were less skewed within the subsets below and above the therapeutic range than in the retrospective study. The proportions of
INRs out of range (after 2 weeks) according to various subsets are shown in Table 2. Overall, among patients randomized to Change 35 (44%) were outside the therapeutic range at follow-up compared to 32 (40%) of those randomized to No change, odds ratio 1.17 (95% confidence interval 0.59-2.30), p = 0.75. When for every baseline INR the follow-up mean INRs were compared in the subset of patients with therapeutic range 2.0-3.0 there were minimal differences up to baseline INR 3.6, above which only 20 patients were recruited (Fig. 4). For the subset with therapeutic range INR 2.5-3.5 the number of patients at every INR was too small for meaningful comparisons. None of the recorded baseline characteristics were associated with the follow-up INR being within the range. In a post-hoc analysis, the median discrepancy of the baseline INR from the closest end of the therapeutic range was 0.3 and 0.4 for patients with follow-up INR within or outside the therapeutic range, respectively (p = 0.062). Adverse events registered during the 2-week observation period were two in the group randomized to No change (ventricular tachycardia triggering the implanted cardioverter/defibrillator; food poisoning) and one in the group randomized to Change (bleeding from penis, baseline INR 2.0, follow-up INR 3.0, therapeutic range 2.5-3.5). Discussion We have in a prospective, randomized study (B) demonstrated that for patients with a stable maintenance dose of warfarin, defined
Fig. 2. Distribution of the qualifying “baseline INRs”, shown as deviations from therapeutic range, in the retrospective cohort (A).
396
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
Table 1 Baseline characteristics for the patients in the prospective randomized study (B). Characteristic
No change of dose
Changed dose
N Male sex, n (%) Age⁎ Indication (main) Atrial fibrillation Mechanical valve, aortic Mechanical valve, mitral Venous thromboembolism Other INR range 2.0-3.0, n (%) Years on warfarin⁎ Months on same dose⁎ Maintenance dose, mg/week⁎ Previous bleeding event Reason for INR deviation Unknown Missed dose Double dose (mistake) Less vegetables in diet More vegetables in diet Antibiotic treatment Other medication change Other reason§ Baseline INR⁎ Below range, INR 2.0-3.0 Above range, INR 2.0-3.0 Below range, INR 2.5-3.5 Above range, INR 2.5-3.5
80 46 (57.5) 70.5 (64-78)
80 49 (61.3) 67.1 (57-79)
Table 2 Proportion of patients outside of therapeutic range after 2 weeks according to randomization (study B). Intensity
Deviation
N
Dosing
Out of range (%)
2.0-3.0
Below range
28 34 38 32 10 6 4 8
No change Increased No change Skipped⁎ No change Increased No change Skipped⁎
36 26 45 47 60 50 50 63
Above range 33 20 13‡ 10 4 66 (82.5) 4.5 (2-10) 7 (5-9) 30 (24.8-37.5) 1
32 24† 13‡ 5 6 66 (82.5) 5 (2-10.2) 7 (5-10.2) 31 (22.5-45.6) 6
43 5 2 12 9 2 6 1
53 3 1 7 9 2 3 2
1.7 (1.6-1.8) 3.4 (3.2-3.7) 1.8 (1.6-2.1) 3.9 (3.8-3.9)
1.7 (1.6-1.8) 3.4 (3.2-3.6) 1.8 (1.6-2.1) 4.0 (3.7-4.1)
None of the differences are statistically significant. ⁎Median and interquartile range. † Two patients had also atrial fibrillation. ‡One patient had also atrial fibrillation. §Other reasons were diarrhea, intake of alcohol or vitamins.
as 3 months without any change, and with an asymptomatic INR result slightly outside of the therapeutic range, there is no appreciable difference between changing a single dose or continuation without any change. In addition we showed in the retrospective study (A) that experienced physicians at an anticoagulation clinic only were prone to order watchful waiting if the INR result was 0.1 INR from the therapeutic range, except for patients with therapeutic range 2.5-3.5 and subtherapeutic INR, where a deviation down to 2.2 appeared acceptable. This resulted in confounding by severity [13] and illustrates the need for randomized, controlled studies. To our knowledge this is the first study where patients were randomized to different dosing in this context. Banet et al described a similar experience in 231 patients on warfarin, where their management had been randomized to anticoagulation service with recommendation to avoid changing the dose or to primary care providers without particular guidelines [14]. In their study only patients with
2.5-3.5
Below range Above range
p-value† 0.61 0.86 1.0 1.0
⁎Skipped or reduced dose. †Pearson's Chi-square with Yates correction, compared to the subset on the preceding line.
INR slightly above the range (INR 3.2, 3.3 or 3.4) were included and there was no uniform management in either group. However, they concluded that in case of an INR of 3.2 or 3.3 those managed at anticoagulation service were more likely to have a follow-up INR within the range than those managed by primary care providers. In our study we had wider margins for eligible excessive INRs but we also required a longer history of stability than in the study by Banet et al (3 months vs. 30 days) [14]. Furthermore, we tried to verify in the telephone interview that the patients did not have a new and continuing risk factor for altered warfarin effect, e.g. long-term medication with amiodarone, antibiotics or long-term change in diet. The proportion of follow-up INRs outside the therapeutic range was higher (40-44%) than the 30% anticipated in the sample size calculation. On the other hand the difference between the groups was minimal. With these proportions, an alpha of 5% and power of 80% we would have required about 2,400 patients per group to show that No change is worse than Change. Due to competing studies on dosing algorithms and on alternative monitoring intervals the recruitment to this study became slower than expected and a decision was therefore made to discontinue recruitment at 160 patients. One of the competing studies recruits patients with at least 6 months of unchanged dose of warfarin. It is possible that without this competition we would have seen a higher proportion of follow-up INRs within the therapeutic range. With the actual sample size of 160 in our study and the proportion of INRs out-of-range in the control group (40%) we ended up with a power of only 47.5% to demonstrate a delta of 15%. There is a power of 80% to demonstrate a delta of 22%, i.e. an increase of out-ofrange INRs from 40% to 62%. The selected time point for follow-up can be discussed. It is obvious that an added or skipped dose of warfarin will affect the INR during the next couple of days, but that is irrelevant in the perspective of this study. Patients with a small deviation of INR are not routinely re-tested after 2-
Fig. 3. Distribution of the qualifying “baseline INRs”, shown as deviations from therapeutic range, in the prospective randomized study (B).
S. Schulman et al. / Thrombosis Research 125 (2010) 393–397
397
modification of the maintenance dose is probably needed but should take into account the magnitude of the deviation. Unnecessary dose changes make the treatment more complex for the patients and should be avoided when no benefit can be demonstrated. When the unexpected INR deviation is more than 0.5 it may be advantageous to perform a change of the maintenance dose. Further studies on all these aspects of warfarin dosing should be encouraged. Conflict of interest statement None. Acknowledgement We are grateful for the assistance by Eric Danzig with the retrospective cohort. References
Fig. 4. Mean “follow-up INR” and standard deviation according to dosing allocation for each INR of “baseline INR” in the prospective randomized study (B). Symbols have been moved slightly sideways in opposite directions for the two randomization arms to avoid overlapping. INR categories with less than 3 data points are not presented.
3 days. In clinical practice the next test would be after 1-2 weeks and our results reflect what is seen then – that approximately 60% of those who were unexpectedly out of range are back in the therapeutic range after 2 weeks, irrespectively of what was done initially. Finally, we used a surrogate outcome, but a clinical event during this short observation period is very unlikely, as illustrated by our adverse events and would have required a completely impractical sample size. Studies on patients with chronic anticoagulation and brief periods of excessive or subtherapeutic INR have shown a low risk of clinically important events and no need for correction with vitamin K or bridging therapy, respectively [2,15]. Due to the limited number of patients in our study we are unable to rule out a 15% difference in the primary outcome and therefore the result is not definitive. However, a large effect can be excluded and therefore we conclude that in stable patients with an INR result up to 0.5 INR from the therapeutic range there is no or only a small difference in the followup INR at 2 weeks if a single dose is changed versus no change at all. This supports the concept of maintaining the same dose without a singledose change that may be confusing for the patient. A repeat INR should be performed after approximately 2 weeks, provided that the reason for the deviation is temporary and removed or that it is unexplained. If the repeat INR still is outside the range in the same direction a small
[1] Ansell J, Hirsh J, Hylek E, Jacobson A, Crowther M, Palareti G. Pharmacology and management of the vitamin K antagonists: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines (8th Edition). Chest 2008;133: 160S–98S. [2] Crowther MA, Ageno W, Garcia D, Wang L, Witt DM, Clark NP, et al. Oral vitamin K vs. placebo in the correction of excessively anticoagulated patients on warfarin: results of a randomized, controlled trial. Ann Intern Med 2009;150:293–300. [3] Hylek EM, Chang YC, Skates SJ, Hughes RA, Singer DE. Prospective study of the outcomes of ambulatory patients with excessive warfarin anticoagulation. Arch Intern Med 2000;160:1612–7. [4] Gunneman T, Ruybalid RL, Jacobson AK, Snell B, Zucker ML, LaDuca FM, et al. Frequent prothrombin time testing reduces inappropriate warfarin dose changes. Thromb Haemost 1999;82:676 (suppl) (abstract). [5] Gage BF, Fihn SD, White RH. Management and dosing of warfarin therapy. Am J Med 2000;109:481–8. [6] Waterman AD, Milligan PE, Banet GA, Gatchel SK, Gage BF. Establishing and running an effective telephone-based anticoagulation service. J Vasc Nurs 2001;19: 126–32. [7] Britt RP, James AH, Raskino CL, Thompson SG. Factors affecting the precision of warfarin treatment. J Clin Pathol 1992;45:1003–6. [8] Dalere GM, Coleman RW, Lum BL. A graphic nomogram for warfarin dosage adjustment. Pharmacotherapy 1999;19:461–7. [9] Horton JD, Bushwick BM. Warfarin therapy: evolving strategies in anticoagulation. Am Fam Phys 1999;59:635–46. [10] Ryan PJ, Gilbert M, Rose PE. Computer control of anticoagulant dose for therapeutic management. Bmj 1989;299:1207–9. [11] Rose AJ, Ozonoff A, Berlowitz DR, Henault LE, Hylek EM. Warfarin dose management affects INR control. J Thromb Haemost 2009;7:94–101. [12] Schulman S. A preliminary study on the feasibility of extending the interval between prothrombin time tests for patients on warfarin therapy. Haematologica 2005;90:132 (abstr 0336). [13] Salas M, Hofman A, Stricker BHC. Confounding by indication: an example of variation in the use of epidemiologic terminology. Am J Epidemiol 1999;149. [14] 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;123:499–503. [15] Clark NP, Witt DM, Delate T, Trapp M, Garcia D, Ageno W, et al. Thromboembolic consequences of subtherapeutic anticoagulation in patients stabilized on warfarin therapy: the low INR study. Pharmacother 2008;8:960–7.