A new regimen for starting warfarin therapy in out-patients
Abstract
Aims Oral anticoagulation is increasingly used in elderly patients with atrial fibrillation to prevent embolic phenomena. The use of anticoagulants in this population is prophylactic rather than therapeutic and so there is no urgency to establish anticoagulation within the desired therapeutic range. The aim of the study was to develop an out-patient regimen for initiation of oral anticoagulation with warfarin which requires only weekly monitoring of the International Normalized Ratio (INR).
Methods The study was undertaken in two phases. In the first phase, factors which predict the final maintenance dosage of warfarin were defined and used to build a decision tree and dosage algorithm. In the second study the algorithm was tested. Patients were given 2 mg warfarin daily for 2 weeks and the INR at this time was used to predict the maintenance dose. Patients then attended for weekly measurements of the INR until steady state had been reached. Dosage adjustments were not made unless the INR was >4.0 or <1.5 for 2 consecutive weeks. The accuracy of the prediction was measured by calculating the mean INR of weeks 6–8 and the number of patients in the target range 2.0–3.0 was determined.
Results One hundred and seven consecutive out-patients (mean age 70 years range 64–86) completed the first study. The age, sex, height, weight, alcohol intake, number of cigarettes smoked, concomitant medication, clinical evidence of right heart failure, liver failure, abnormalities in liver enzyme estimations, baseline INR and INR after 2 weeks of 2 mg warfarin daily were used in a polytomous logistic regression analysis with stepwise inclusion of factors to determine which factors influenced the eventual maintenance dosage of warfarin. The INR after 2 weeks of 2 mg warfarin therapy predicted 70% of the variability of the maintenance dose. Of other factors only the sex of the patient had a large enough effect to be included in the prediction algorithm. One hundred and six patients (mean age 71 years range 50–85 years) completed the second study. Only one patient needed a dose adjustment in the first 2 weeks of warfarin 2 mg daily (INR 4.4). Overall, 60% patients were in the narrow target range (INR 2.0–3.0) at steady state. In five patients the INR was >4.0 at any visit after the second week and needed dosage adjustment. In four patients the INR was <1.5 at steady state.
Conclusions We have developed a method of predicting the maintenance dose of warfarin in an elderly population based on the INR after 2 weeks of warfarin 2 mg daily, and the sex of the patient. This is a safe and convenient way of initiating warfarin therapy as an out-patient which requires only weekly INR checks.
Introduction
Anticoagulation with warfarin is increasingly indicated. It is of proven benefit in the management of thromboembolism and may be used as prophylaxis against thromboembolism in patients following acute myocardial infarction [1] and in those with atrial fibrillation [2]. The use of oral anticoagulation in patients with atrial fibrillation has increased rapidly over the last few years following the publication of several clinical trials which have shown its benefit [2, 3]. This dysrhythmia increases in incidence with age and it has been estimated that it occurs in about 4% of patients over 74 years and 8% of octogenarians [4, 5]. Anticoagulation in the elderly is potentially hazardous and careful monitoring is important to maintain a positive benefit:risk ratio [6].
Practical anticoagulation
Warfarin is the drug of choice for oral anticoagulation therapy. Warfarin has a long-terminal elimination half-life and its effect is dependent upon the suppression of the synthesis of normal clotting factors such that there is a delay before the clinical effect is seen. Most regimens for anticoagulation involve loading with a large dose of warfarin and then adjusting the maintenance dosage on a daily basis by comparison with the prothrombin time of the patient's blood against a standard. An adjusted ratio, the international normalized ratio or INR, is then derived which is used to select dosage changes. Guidance in choosing the appropriate daily dosage of warfarin in response to the INR can be obtained from treatment tables [7] or computer derived decision support tools [8]. Eventually, when the INR is stable in the target range the patient continues treatment with the maintenance dosage. This process of dose titration usually takes 5–7 days.
When anticoagulation is required for prophylactic reasons rather than therapeutic reasons there is less need for haste. Moreover, anticoagulation cannot always be started with the patient as an in-patient where blood monitoring is easy. The ability to initiate anticoagulation in the out-patient department safely with only weekly attendance for blood monitoring would have great benefits for the health service and for patients.
The study had two aims. Firstly, to identify factors which could predict the maintenance dosage of warfarin in an elderly population and secondly, to use the factors to develop a decision algorithm which could be used to predict the eventual maintenance dosage of warfarin. In the second part of the study we tested the algorithm.
Methods
The study was designed in two phases. The first phase was to establish which factors influenced the maintenance dose of warfarin for the narrow target range of INR 2.0–3.0. From this database an algorithm was produced to predict the maintenance dose and this algorithm was tested prospectively in the second phase. For both phases of the trial, eligible patients were out-patients who required nonurgent oral anticoagulation (not requiring intravenous heparin initially). No age limits were set for entry into the studies and there were no exclusion criteria at entry in addition to those for routine anticoagulation. Patients gave informed consent and were supplied with a patient information sheet describing the study aims and giving important background information about the risks and benefits of oral anticoagulation treatment. The study was approved by the hospital ethics committee.
One hundred and seven (67 males; 40 females) consecutive out patients of age range 64–86 years (mean=69.9 years) completed the first phase of this study between August 1994 and May 1995. One hundred and six (71 males; 35 females) consecutive out-patients of age range 50–85 years (mean=70.7 years) completed the prospective trial between July 1995 and May 1996.
Indications for anticoagulation are shown in Table .
Study 1
At baseline (week 0) demographic details, concomitant medication, alcohol, and tobacco use were recorded. Patients were examined for clinical evidence of heart failure and liver disease; height and weight were measured.
Protocol At week 0, the baseline INR and liver function tests were recorded, and 2 mg warfarin was given for 2 weeks.
At week 1, the patient returned to the clinic and the INR was checked but the dosage of warfarin was continued at 2 mg day−1.
At week 2, the patient returned to the clinic and the INR was checked and warfarin dosage was adjusted according to the following rules:
3 mg if INR <2.0
2 mg if INR 2.0–3.0
1 mg if INR >3.0
At week 3, the patient returned to the clinic and the INR was checked. If the INR was in the range 2.0–3.0 the dosage was continued unchanged. If the INR was >3.0, the dosage of warfarin was reduced by 1 mg day−1; if the INR was <2.0 the dosage of warfarin was increased by 1 mg day−1.
The patient returned to the clinic at weekly intervals for INR checks and if necessary, warfarin dosage adjustments by 1 mg steps. At week 6 if the INR was not yet stable in range 2.0–3.0 then 0.5 mg dosage adjustments were made.
The end point of the study was reached when the INR had been in the range 2.0–3.0 for 3 consecutive weeks and for the last 2 weeks the dosage of warfarin was not changed. If any of the weekly INR results were missing, the patient was excluded from the study.
The results from this study were used to identify the predictors of warfarin maintenance dosage. The data were analysed using polytomous logistic regression with stepwise inclusion of factors which could be used in conjunction with the INR at 2 weeks to predict the maintenance dose of warfarin. The following factors were used: age, sex, height, weight, alcohol intake, number of cigarettes smoked, concomitant medication, clinical evidence of right ventricular failure, clinical evidence of liver failure, abnormalities in liver enzyme activities and baseline INR.
From this analysis an algorithm was derived which was subsequently tested prospectively.
Study 2
Baseline INR and demographic details were recorded.
All patients were given 2 mg day−1 of warfarin for 2 weeks but returned to the clinic weekly for an INR measurement. On the basis of the INR result after 2 weeks of warfarin 2 mg day−1, the predicted maintenance dose was prescribed according to the algorithm (see Table 4). Patients continued to attend on weekly occasions for INR measurements, but the dose of warfarin was not changed unless the INR >4.0. If the INR was >4.0, warfarin was omitted for 2 days and then restarted at a dose 1 mg lower. If the INR was <1.5 for 2 consecutive weeks the dose of warfarin was increased by 1 mg.
Logistic regression techniques were again used with the same independent variables to determine whether the demographic data could be used to predict:
1 those patients who would have an INR >4.0 at any stage after week 2
2 those patients with an average INR >3.0 over weeks 6–8
3 those patients with an average INR <2.0 over weeks 6–8
Efficacy of anticoagulation
The efficacy of anticoagulation with our new starting regimen was tested by counting the number of patients with an INR in the narrow therapeutic range chosen for our study (INR=2.0–3.0). The mean INR at weeks 6,7,and 8 was used to determine the success of the prediction. If any of these values were missing the patient was excluded from the study. We also used the method described by Rosendaal et al. [9] which allows a calculation of the time spent within the target range for weeks 6–8.
Results
Study 1
The demographic details of the population in the first study are shown in Table 2. The INR measured at week 2 and the final steady state warfarin dose were positively skewed. There was a close correlation between week 2 INR and warfarin dose with a Spearman rank correlation suggesting that 70% of the variability of the final warfarin dose was predicted by the INR at 2 weeks. Because of the discrete nature of the data and the likely nonlinear link between INR and warfarin dose, polytomous logistic regression with stepwise inclusion of all the possible predictive factors was used to determine predictions of warfarin maintenance dose. Three further parameters had significant influences on the final dose of warfarin (see Table 3); these were sex, clinical evidence of right heart failure and use of angiotensin converting enzyme inhibitors. However, in designing a simple algorithm for use in clinical practise, we omitted the effects of angiotensin converting enzyme inhibitor drugs and the presence of clinical evidence of right heart failure. Of other drugs known to interfere with warfarin one patient in this group was taking amiodarone and phenytoin. Thus, the algorithm that we used in the second study, only included the INR at week 2 after 2 weeks treatment with 2 mg warfarin and the sex of the patient (see Table 4) and used only 1 mg increments.
Study 2
In the prospective trial 95% of weekly INR measurements were recorded. There were no adverse events related to high or low levels of anticoagulation.
One patient had an INR of 4.4 after 2 weeks treatment with 2 mg warfarin. This patient also had clinical features of right heart failure but no other distinguishing features. Five patients had INRs >4.0 at any visit after week 2 and required dose adjustment according to the protocol (see Table 5). Logistic regression failed to find any significant predictors of this high INR from the independent demographic variables. Twenty-one patients had average INRs >3.0 at weeks 6–8. Again none of the demographic variables identified at entry into the study was able to predict the outcome in these patients. For patients with INRs <2.0, however, there was a significant effect of younger age and higher week 2 INR. Patients over the age of 75 years were four times less likely to be under anticoagulated and there was a tendency for those patients with high INRs at week 2 to then receive too low a maintenance dose of warfarin. This indicates that this algorithm is more likely to under anticoagulate younger patients, and perhaps is best suited to an elderly population.
Table 6 shows the percentage time spent in different INR ranges for the weeks 6–8. Within the population of 106 patients <1% of time was spent with an INR of >4.0. The majority of time was spent in the range 2.0–4.0 during this period although there was a substantial time in the under anticoagulated range (<2.0).
Discussion
We have developed a decision algorithm for out-patient anticoagulation with warfarin in an older population. The algorithm was designed to initiate anticoagulation for prophylactic rather than therapeutic purposes, and in particular for the large group of elderly patients with atrial fibrillation. The over-riding concern was safety and dosing was designed to be conservative and to avoid over anticoagulation since this is the cause of most morbidity and mortality in this setting. We chose a very narrow range at which to aim (INR 2.0–3.0) as recommended recently [10] yet despite this 60% of our patients had INRs in this range and 50% of time was spent in this range for the population as a whole. Considering the target range recommended for non rheumatic atrial fibrillation by some authors [11] of 2.0–3.9, about 80% of our patients had INRs in this range and the population spent 62% of time in this therapeutic range. Our algorithm is conservative with no patient recording an INR >4.5 and only <1% of time was spent with an INR >4.0.
Although the percentage time spent in the target range would appear to be low (49%) it is similar to that achieved after 6 months anticoagulation in a large clinical trial of anticoagulation after acute myocardial infarction [12]. These authors found that patients spent <50% of time in the target range [2.8–4.8] within the first month of starting anticoagulation and 66% between months 1–3. This range was twice as wide as the one that we set as our target. They also found that the majority of patients who were not in the target range were under anticoagulated in the first 6 months of treatment. Even in the context of a clinical trial of anticoagulation following myocardial infarction, the authors accepted stable anticoagulation when 70% of time was spent in the target range. We believe our regimen achieves similar results.
Cautionary notes
We identified in study 1 that two other factors influenced the final dosage of warfarin. Concomitant coprescription of an angiotensin converting enzyme inhibitor drug increased the dosage of warfarin required whereas the presence of right heart failure reduced the dosage. The frequency of these cofactors was small so we made a decision not to include them in the algorithm. Our study population also included only two patients who were taking drugs (amiodarone and phenytoin) known to interfere with warfarin handling. It is important therefore to use this protocol with caution in patients in right heart failure and in those taking concomitant drugs known to interfere with warfarin. Indeed, in the second study the only patient who was potentially over-anticoagulated was a patient with right heart failure. Since the algorithm is conservative it is also inappropriate for therapeutic anticoagulation when it is important to establish anticoagulation in the therapeutic range as soon as possible.
Our data and others [13] show that elderly patients are more sensitive to warfarin, and although we had no patients with bleeding during the study we cannot exclude the possibility of serious over anticoagulation; however, since bleeding risk increases with an INR >4.0 this risk should be small [14, 15]. Our initial intention was to devise a system where patients were seen only at 2 and 6 weeks. The frequency of INR measurements above 4.0 leads us to advise weekly visits until steady state has been achieved so as to avoid such over-anticoagulation. Moreover, the population involved are elderly and regular visits to the anticoagulation clinic allows important time for education of patients about the risks, benefits and precautions necessary when taking warfarin. It also provides an opportunity to check compliance.
We performed this study in a hospital out-patient anticoagulation clinic although we believe that the regimen is suited for out-patient anticoagulation in a community setting and supervised by general practitioners. Logistically, attendance at an out-patient clinic on a weekly basis is considerably easier and cheaper than using alternative algorithms which require daily INR measurements [7] and so often necessitate hospital admission.
