The Journal of Informed Pharmacotherapy 2002;8:203-205.
Download a PDF version of this article
Reviewer: Stephen Shalansky, Pharm.D., FCSHP
Reviewer's email address: firstname.lastname@example.org
Reviewer's profession/specialty: Research Coordinator, Pharmacy Department/Cardiovascular Pharmacotherapy
Hylek EM, Regan S, Go AS, Hughes RA, Singer DE, Skates SJ. Clinical predictors of prolonged delay in return of the international normalized ratio to within the therapeutic range after excessive anticoagulation with warfarin. Ann Intern Med 2001;135:393-400. PubMed Citation
The objective of this study was to identify predictors of the rate of INR normalization after excessive warfarin anticoagulation. The authors suggest this information would be useful for identifying patients who should receive phytonadione (vitamin K1); namely, those patients at high risk for prolonged excessive anticoagulation for whom aggressive intervention is warranted. This information might also be used to identify patients who could safely forgo treatment with phytonadione, thus avoiding the possibility of increased thromboembolic risk through over-correction of INR.
These authors have published previous research on anticoagulation including a widely quoted article identifying acetaminophen as a risk factor for excessive anticoagulation with warfarin (1). For the current study, they have used a similar study design and drawn upon the same patient population, namely those patients attending their anticoagulation clinic. All patients attended the clinic between 1993 and 1998, had an INR greater than 6.0, two subsequent warfarin doses held, were interviewed within 24 hours of their elevated INR, and had a follow-up INR on the second calendar day. Patients whose INR was <4.0 on day 2 were compared to those whose INR was >4 on day 2. Variables evaluated as potential risk factors for INR >4 included age, warfarin dose, recently started medications, concurrently illness, decreased oral intake, confusion about their warfarin dose, non-adherence, hospitalization, and increased alcohol intake. Data was obtained through review of clinical and hospital records.
1. Was assignment of patients randomized?
No. This was a retrospective, cohort study.
Of the 7279 managed in the study centre during the study period, 1995 had at least one INR >6.0. Of these, 639 met the inclusion criteria. Six patients were excluded: five because they had received phytonadione within the first 48 hours and one with hepatitis. Of the 633 patients included in the analysis, nursing notes on alcohol consumption, oral intake and new medications were missing for 21 patients (3%). There was no explanation of how the missing data was handled in the analysis.
1. How large was the treatment effect and how precise was the estimate of this effect?
Of the 633 study patients, 232 (37%) had an INR >4.0 after two doses of warfarin were withheld (i.e. on day 2). Independent risk factors for elevated INR on day 2 were older age (OR 1.18 per decade, 95% CI 1.01 – 1.38, p=0.04), higher index INR (OR 1.25 per unit increase, 95% CI 1.14-1.37, p<0.001), decompensated congestive heart failure (OR 2.79, 95% CI 1.30-5.98, p=0.009), and active cancer (OR 2.48, 95% CI 1.11-5.57, p=0.03). Higher weekly warfarin dose was associated with a lower odds of an elevated INR on day 2 (OR 0.87 per 10 mg increase, 95% CI 0.79-0.97, p=0.009). The authors also present results of model-generated probabilities of the combined effects of these risk factors with several hypothetical examples. For example, a 60 year old patient taking 57.5 of warfarin per week has a model-generated probability of 14% for an INR >4.0 on day 2, while an 80 year old taking 42.5 mg/week has nearly three times the risk (40%).
2. Were all clinically important outcomes considered?
Warfarin anticoagulation is highly variable and can be influenced by a wide variety of factors. There are hundreds of potential pharmacokinetic and pharmacodynamic variables to consider. Therefore, it would be neither technically feasible nor statistically appropriate to attempt to account for all potentially important variables in a regression model where the objective is to identify risk factors for excessive anticoagulation.
Hylek et al. have included a relatively large number of patients with outcome of interest in their analysis allowing them to evaluate a considerable number of variables without compromising model stability (a maximum of 10 patients with events per variable has been proposed. (2) However, there is sparse information about the statistical methods use to build and test the model. Therefore, there is some doubt whether accepted steps in regression model analysis; such as testing for interactions between variables, testing for co-linearity, and considering outliers; were carried out. (3)
This study was carried out at one site, and the model was not validated in another patient sample, therefore the risk factors identified here may not be applicable to patients in other settings. Furthermore, data was collected exclusively from historical records raising questions about the accuracy and thoroughness of the data used in the analysis. For example, there is no explanation of how information about medication use and adherence was documented in the clinic records, or the period over which patients were expected to recall relevant exposures. Patient recall is often incomplete or inaccurate, which may have lead to an underestimation of the risk associated with some of the variables evaluated (e.g. alcohol consumption or specific medications).
It is interesting to consider that Hylek’s earlier study on risk factors for excessive warfarin anticoagulation identified “newly started potentiating medication” as a strong risk factor (OR 8.5, 95% CI 2.9-24.7, p<0.001), while medications did not appear to have a significant independent effect in this more recent study.
Finally, other potentially important risk factors were not considered including adherence with concurrent potentiating medications or the use of alternative medicines.
Even with the limitations of this study outlined above, the risk factors identified are likely important in many patients taking warfarin, and should be considered when evaluating the need for phytonadione administration in patients with elevated INR results. The authors suggest that elderly patients taking low doses of warfarin (<15 mg/week), especially those with cancer or decompensated congestive heart failure, should receive early phytonadione administration to limit the chance of major hemorrhage. Conversely, they claim that the risk of intervention with phytonadione probably outweighs the benefit for younger patients with warfarin doses exceeding 50mg per week. However, an accompanying editorial suggests that the most recent guidelines for phytonadione administration limit the chance of over correction of the INR 4.
Given the complexity of warfarin pharmacodynamics, it is particularly important to evaluate each patient’s need for phytonadione based on their individual characteristics and exposures. The risk factors identified in this study may not be the most important issues to consider for many patients encountered in your clinical practice.
Copyright © 2002 by the Journal of Informed Pharmacotherapy. All rights reserved.