| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Clinical Pharmacy Research Team (Dr. Delate), Clinical Pharmacy Anticoagulation Services (Drs. Witt and Jones, and Mr. Senser), and Clinical Pharmacy Services-Nephrology (Dr. Bhardwaja), Kaiser Permanente Colorado, Aurora, CO.
Correspondence to: Daniel M. Witt, PharmD, PharmD, FCCP, Kaiser Permanente Colorado, 16601 E Centretech Pkwy, Aurora, CO 80011; e-mail: dan.m.witt{at}kp.org
Abstract
Background: Elevated international normalized ratio (INR) values have been linked to bleeding complications; however, elevated INR values are not always physiologic and can be falsely increased. This study describes the rate of falsely elevated INRs and characteristics predictive of falsely elevated INRs.
Methods: This cross-sectional study was conducted among adult patients receiving anticoagulation therapy monitored by a centralized anticoagulation service during January 2000 through December 2004 (n = 29,536). Prevalence rates of all elevated (ie, value 
10), falsely elevated, and truly elevated INRs were calculated. Multivariate logistic regression was performed to identify predictors of falsely elevated INRs among elevated INRs.
Results: Of the 556,998 INRs included in the analysis, 793 INRs (prevalence, 0.14%; 95% confidence interval [CI], 0.10 to 0.19%), 53 INRs (prevalence, 0.01%; 95% CI, < 0.01 to 0.03%), and 740 INRs (prevalence, 0.13%; 95% CI, 0.09 to 0.18%) were elevated, falsely elevated, and truly elevated, respectively. The strongest independent predictor of a falsely elevated INR was a patient undergoing hemodialysis at the time of the elevated INR (adjusted odds ratio, 9.60; 95% CI, 4.96 to 18.58; p < 0.001). A low target INR was the only other factor found to be an independent predictor of a falsely elevated INR.
Conclusions: Although INR values 10.0 occur infrequently, patients presenting with such values can present a challenge to the anticoagulation provider. Anticoagulation providers should be particularly vigilant for falsely elevated INRs when monitoring patients undergoing hemodialysis.
Key Words: anticoagulation • bleeding • hemodialysis • international normalized ratio
Oral anticoagulation therapy with warfarin is monitored routinely using international normalized ratios (INRs). Elevated INR values have been linked to bleeding complications; thus, accurate INR monitoring is required during anticoagulation therapy.1 However, elevated INR values are not always physiologic and can be falsely increased by factors including evacuated blood collection tube fill volume and citrate concentration.23 Falsely elevated INRs do not confer increased bleeding risk and in fact may lead to unnecessary anticoagulant therapy reversal and subsequent risk for thromboembolic complications.4
Anticoagulation providers at the Kaiser Permanente Colorado (KPCO) Clinical Pharmacy Anticoagulation Service (CPAS) have noted that some INRs are initially superelevated (INR 10.0) but within the therapeutic range (INR 2.0 to 3.0) when rechecked the following day without therapeutic intervention (ie, a falsely elevated INR value). This phenomenon seems to occur more commonly under certain clinical circumstances, most notably in patients undergoing hemodialysis and in patients whose blood is collected using pediatric evacuated collection tubes.
It has been demonstrated previously that inadequate filling of pediatric collection tubes increases the potential of falsely elevated INR values.3 Preliminary investigation of a sample of hemodialysis-associated falsely elevated INR episodes at KPCO revealed that blood for these INR measurements was frequently drawn prior to hemodialysis via a central venous catheter (CVC), whereas the therapeutic repeat samples were drawn via a peripheral vein. CVCs are prone to complications, including thrombosis, necessitating the need for periodic instillation of heparin into the lumen of the line.5 It has been hypothesized that heparin contamination of blood samples collected from CVCs may compromise INR reliability.67 A literature review5 found limited research detailing the accuracy of INR values obtained via CVCs.
Adjustment of warfarin dosing based on falsely elevated INR values increases the risk of treatment failure (thrombosis).3 Description of the patient characteristics associated with falsely elevated INRs in anticoagulation patients will help anticoagulation providers recognize patients at risk for this phenomenon. The purpose of this investigation was to describe the rate of truly elevated and falsely elevated INRs in a diverse, real-world anticoagulation patient population and to identify patient characteristics predictive of falsely elevated INRs among anticoagulated patients with an elevated INR.
Materials and Methods
Setting
This cross-sectional study was conducted at KPCO, a group model health maintenance organization providing integrated medical care to approximately 430,000 patients in the Denver-Boulder metropolitan area. The study was reviewed and approved by the KPCO Institutional Review Board.
The KPCO CPAS provides comprehensive services for all KPCO patients requiring anticoagulation therapy and currently monitors > 6,800 patients. Working collaboratively with the referring physician, CPAS clinical pharmacists initiate anticoagulation therapy, order relevant laboratory tests (including INRs), adjust anticoagulation medications as necessary, and refill anticoagulation medications. Guidelines have been implemented at KPCO that standardize evacuated blood collection tube fill volume and citrate concentration during blood sampling for INR measurements. All INRs included in this evaluation were derived from venous blood samples processed by in-house KPCO laboratories, including specimens collected from patients undergoing hemodialysis.
Patients receiving warfarin anticoagulation therapy monitored by CPAS Pharmacists and having had an INR measurement during the period of January 1, 2000, through December 31, 2004 (study period) were eligible for inclusion. Patients < 18 years of age during the study period were excluded.
Data Collection
The KPCO integrated medical, pharmacy, and laboratory records system was queried to identify the dates and values of all INR measurements performed during the study period and the age of the patients receiving the measurement. Health record numbers of eligible patients were used to query the CPAS anticoagulation database (Dawn-AC; 4S Systems, Ltd; Cumbria, UK) to identify gender, primary indication for anticoagulation therapy, target INR range, and hemodialysis status and anticoagulation inception dates. Eligible patient health record numbers were then used to query the integrated records system again to identify the reception of vitamin K or other medications that may alter INR values during the study period (a list is given in the Appendix). Nursing home status and KPCO membership begin and end dates were also recorded.
Outcomes
The primary outcomes were the prevalence rates of all elevated INRs, truly elevated INRs, and falsely elevated INRs, respectively, in an anticoagulation patient population. Additional analyses were preformed to identify patient characteristics predictive of falsely elevated INRs among patients with an elevated INR. An elevated INR was defined as an INR 10.0. A falsely elevated INR was defined as an initial INR 10.0 but 
3.0 when another INR measurement was performed the following day (without administration of vitamin K). A truly elevated INR was defined as an initial INR 10.0 without a follow-up INR the following day (with or without administration of vitamin K) or a follow-up INR > 3.0 the following day (with or without administration of vitamin K) [Fig 1
].
|
Patients with an elevated INR were assigned to either the falsely elevated or truly elevated INR cohorts (there were no patients with both a falsely elevated and a truly elevated INR). Prevalence rates of all elevated, falsely elevated, and truly elevated INRs were calculated as the count of the respective INR divided by the total count of eligible INRs performed during the study period and reported as counts per 10,000 INRs performed. In addition, 95% confidence intervals (CIs) around the prevalence values were calculated.
Medications that could potentially elevate an INR (shown in Appendix) were assessed to confirm that the patient had a supply of the medication in possession at the time of the elevated INR. For example, a patient who received a 4-day supply (indicating that the medication was intended to be used for 4 days) of amoxicillin 2 days before an elevated INR was measured would have been flagged as having had a medication that potentially could elevate an INR in possession at the time of the elevated INR. Time elapsed (in days) since initiation of warfarin therapy was used as a marker of patient familiarity with anticoagulation therapy and its management. This was based on the assumption that patients more recently initiated on anticoagulation therapy would be at higher risk of abnormal INR values due to their lack of familiarity with the various requirements of anticoagulation therapy. Nursing home status at the time of the elevated INR was included as a characteristic because blood sampling would have been performed by non-KPCO phlebotomists, thus allowing for variance in sample handling procedures. The primary indications for anticoagulation therapy were collapsed into three categories (ie, atrial fibrillation, venous thrombosis, and other) for the predictive modeling due to the relatively limited representation of patients with some indications and the subsequent inability of the model to reach convergence with empty cells.
Distributions of interval-level and ratio-level variables were assessed for normality. Patient characteristics were reported by cohort as means and SDs or proportions. Due to some patients having had multiple elevated INRs (ie, observations) during the study period, both time-varying (eg, age at the time of the elevated INR) and non–time-varying (eg, gender) characteristics were present in the data. Thus, to adjust for interobservation correlations during the assessment of differences between the cohorts, logistic regression (PROC SURVEYLOGISTIC, SAS v9.1; SAS Institute; Cary, NC) was fit by maximum likelihood, but the variances were adjusted for clustering of the observations.8 Clustering was set at the individual patient level. Multivariate logistic regression was performed with PROC SURVEYLOGISTIC to identify predictors of falsely elevated INRs among elevated INRs. A fully saturated model was first constructed with patient characteristics and their interactions. As no interaction terms were found to be significant predictors, only the main effects were input into the final model.
As this is an exploratory investigation, an 
level of 0.05 was utilized to limit the potential for type I errors. In addition, no a priori power analysis was performed since the vast majority of patients undergoing oral anticoagulation therapy during the study period were eligible for inclusion in the analysis.
Results
A total of 558,244 INRs were identified during the study period. Of these, 1,246 INRs belonging to one patient who was < 18 years of age at the time of the INR and other patients for whom no patient identifier was recorded were excluded. Thus, 556,998 INRs, representing 29,536 patients, were included in the analysis. A total of 793 elevated INRs, representing 624 patients, were identified for an elevated INR prevalence of 14 per 10,000 INRs performed (95% CI, 0.10 to 0.19%). Of the 108 potential falsely elevated INRs, 53 INRs (6.7% of all elevated INRs), representing 46 patients, were confirmed to be falsely elevated for a prevalence of 1/10,000 INRs performed (95% CI, < 0.01 to 0.03%). The remaining 740 elevated INRs, representing 578 patients, were determined to be truly elevated for a prevalence of 13/10,000 INRs performed (95% CI, 0.09 to 0.18%).
Patients with a falsely elevated INR were more likely to have had an INR target < 3.0 (p < 0.001) and been receiving hemodialysis (p < 0.001) at the time of their elevated INR compared to patients with a truly elevated INR (Table 1 ). Conversely, patients with a truly elevated INR were more likely to be residing in a nursing home (p = 0.001). In addition, differences in proportions of patients with a valve disorder (p = 0.014) and coronary artery disease (p = 0.011) primary indication for anticoagulation therapy were present between the groups. Differences in time from initiation of warfarin therapy to the elevated INR and the proportion with any medication likely to increase an INR value were not significant. However, patients with truly elevated INRs were more likely to have been prescribed antiinfective agents than patients with falsely elevated results (p = 0.022) [Table 2 ].
|
|
|
|
The results of this descriptive investigation demonstrate that INRs 10.0 occur infrequently in anticoagulated patients managed by a centralized anticoagulation service (14/10,000 INRs performed, or 0.14%). In addition, the results confirm that although falsely elevated INRs do occur during routine anticoagulation, they are extremely rare (1/10,000 INRs performed, or 0.01%). However, falsely elevated INRs comprised nearly 7% of all INRs 10.0 that occurred during the study interval, suggesting that anticoagulation providers should carefully evaluate very high INR results before taking aggressive action to reverse anticoagulation therapy.
The strengths of our study include the inclusion of over one-half million INR measurements in the analysis and the identification of strong predictive identifiers for potential false INR elevations. The results of our investigation should be generalizable to other settings given the large, diverse anticoagulation population from which the data were collected.
In our bivariate analysis, we found that patients receiving hemodialysis and those with target INRs < 3.0 were at higher risk for falsely elevated results. Conversely, patients residing in nursing homes more commonly had truly elevated results. The results of the multivariate analysis indicated that being on hemodialysis at the time of the elevated INR is an independent and strong risk factor for having a falsely elevated INR (risk increased nearly tenfold).
As anticoagulation therapy is used in approximately 6% of hemodialysis patients, reliable INR results are integral to the care of these patients.9 Blood for INR testing in patients on hemodialysis is frequently collected directly from the dialysis access site in order to minimize additional patient discomfort and inconvenience from peripheral venipuncture. The results of this study indicate that in some cases, this practice may paradoxically increase discomfort and inconvenience secondary to the need for repeat blood sampling from a peripheral vein to rule out the possibility of erroneous results. It appears that blood collected via heparinized CVCs is particularly prone to producing falsely elevated INR results. Unfortunately, available studies5-7,10–14 disagree about the best approach to obtaining blood for INR testing in this clinical situation. Some authors61011 advocate discarding various volumes of blood from the CVC just prior to collecting the sample to be used for INR testing. Drawing blood directly from the dialysis circuit bloodline arterial sample port has also been evaluated.5 Others authors7121314 suggest that peripheral venous sampling is the only valid method of obtaining reliable INR results. Attempts at KPCO to institute uniform blood collection guidelines for INR testing in patients undergoing hemodialysis have met with mixed results. Therefore, we suggest that blood samples for INR testing in patients undergoing dialysis be collected from a peripheral vein. Before aggressive measures are taken to reverse anticoagulation therapy, very high INR results derived from blood drawn from dialysis access sites, particularly from CVCs, should be confirmed as soon as possible via a peripheral sample.
The finding that the likelihood of a falsely elevated INR is increased in patients with INR targets < 3.0 is of limited clinical utility, as the vast majority of patients receiving warfarin anticoagulation therapy fall into this category. Other information contained in this study may provide more practical utility in assisting anticoagulation providers as they develop therapeutic plans for patients with very high INRs. We have previously demonstrated that inadequate filling (< 90% of total fill volume for tubes containing 3.2% citrate concentration) of pediatric collections tubes results in falsely elevated INR results.3 Therefore, anticoagulation providers should routinely verify what type of collection tube was used before making therapeutic plans for patients with very high INRs. In the absence of bleeding, vitamin K should be withheld until an elevated INR derived from blood drawn in a pediatric tube is confirmed through repeat testing using an adequately filled adult collection tube (at least 60% of total fill volume for tubes containing 3.2% citrate concentration).2
Irregularities in sample collection were noted in some cases of falsely elevated INRs in our study. Examples included combining the contents of two underfilled collection tubes and excessively traumatic venipuncture. Suspicion of falsely elevated INR results should also be considered in patients with previously stable INR control who present with very high INR results in the absence of any circumstance that would explain the elevated result (eg, interacting medications, intercurrent illness, recent alcohol consumption, changes in dietary vitamin K consumption). Consideration should be given to confirming unexplained elevated INR results associated with any of the aforementioned circumstances via repeat sampling.
Identification of the problem of falsely elevated INR results became apparent in our health-care system following the implementation of our centralized CPAS. Clinicians with extensive experience in the nuances of warfarin therapy management may be more likely to recognize circumstances where falsely elevated INR results have occurred.4
The retrospective design of our study exposed our results to several limitations. We examined a limited set of potential predictors due to the lack of availability of or difficulty obtaining some variables (eg, smoking status, aspirin usage). Also, we were unable to assess the predictive value of blood samples when collected in pediatric tubes, short-filled tubes were combined together, and a traumatic venipuncture occurred during collection because this information was not noted routinely during INR blood sample collection. Furthermore, some variables may not have been identified as significant predictors of falsely elevated INR results due to the limited number of patients included in the multivariate analysis. Although overall INR control for patients enrolled in CPAS was not included as a variable in the analysis, the percentage of time patients spent in the therapeutic INR range remained very consistent throughout the study interval (range, 61.4 to 63.7%).
We used a relatively high cut-off point for defining elevated INRs. Our definition was chosen based on the anecdotal observation that falsely elevated INRs tend to be very elevated, often above the upper limit of the instrument used to perform the INR testing. In our laboratory, that upper INR limit has been in the range of 10.8 to 12.9 during the study period. Our definition of elevated INR was also chosen because vitamin K administration is generally warranted for truly elevated INRs 10.0.15 As the administration of vitamin K and subsequent reversal of anticoagulation in patients without truly elevated INRs can lead to iatrogenic thrombotic complications,3 we believe that our 10.0 cut-off point was justified clinically. Additionally, we placed all patients who received vitamin K immediately after the elevated INR was reported into the truly elevated cohort. This may have resulted in some elevated INRs being inaccurately coded as truly elevated, especially in nursing homes where vitamin K is routinely on hand. Thus, patients in this setting with an INR 10.0 could have received vitamin K even though the INR was falsely elevated, leading to some inaccuracy in the reported estimates of falsely elevated and truly elevated INRs.
Although INR values 10.0 occur relatively infrequently, patients presenting with such values present a significant challenge to the anticoagulation provider. Patients with high INRs and evidence of significant bleeding should be treated aggressively with vitamin K administration and administration of exogenous clotting factors if necessary.15 For asymptomatic patients, the anticoagulation provider should carefully consider the possibility of false INR elevation as part of the therapeutic plan. Confirming the accuracy of the elevated INR result may prevent the inappropriate reversal of anticoagulation therapy and the attendant potential for thrombotic complications or refractoriness to continued warfarin therapy. Anticoagulation providers should be particularly vigilant regarding elevated INRs collected from patients undergoing hemodialysis or into pediatric tubes.
Appendix
Medications that may alter INR values during the study period include antiinfectives (eg, trimethoprim/sulfamethoxazole, metronidazole, erythromycin, clarithromycin, ciprofloxacin, fluconazole, itraconazole, ketoconazole); sulfinpyrazone; amiodarone; nonsteriodal antiinflammatory drugs; antithyroid agents (eg, propylthiouracil, methimazole); thyroid hormones (eg, levothyroxine, liothyronine, desiccated thyroid); barbiturates; carbamazepine; danazol; diflunisal; disulfiram; antidepressants (eg, fluoxetine, paroxetine, sertraline, fluvoxamine); phenytoin; propafenone; stomach ulcer/acid reducing agents (eg, cimetidine, omeprazole); and lipid-lowering agents (eg, fibrates and statins).
Footnotes
Abbreviations: CI = confidence interval; CPAS = Clinical Pharmacy Anticoagulation Service; CVC = central venous catheter; INR = international normalized ratio; KPCO = Kaiser Permanente Colorado; OR = odds ratio
The authors have no conflicts of interest to disclose.
Received for publication September 5, 2006. Accepted for publication September 25, 2006.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
