Correlation of Point-of-Care International Normalized Ratio to Laboratory International Normalized Ratio in Hemodialysis Patients Taking Warfarin

Materials and Methods

All chronic IHD patients receiving warfarin anticoagulation to a therapeutic goal of 2 to 3 or higher for at least 1 mo were approached for study participation. Mayo Foundation has two hemodialysis centers that are staffed by nurses, physicians, and practitioners working within the division of Nephrology at Mayo Clinic, Rochester. The Mayo Clinic Institutional Review Board approved this study.

In this prospective comparison of paired patient data, blood samples were collected serially within 10 min immediately before hemodialysis circuit accession and heparin initiation. Each patient provided at least six sets of blood samples for INR, including the standard laboratory draw as well as POC INR samples from finger stick and from their IHD access. Serial blood samples obtained by standard finger stick and dialysis access site were analyzed immediately using the CoaguChek S device (Roche Diagnostics, Indianapolis, Indiana) (6). Dialysis access included either a central venous catheter (typically an Ash-split catheter), AV fistula, or graft. Whole blood samples (5 ml) for plasma INR assessment were collected into 3.8% sodium citrate (1:9 dilutions). Plasma INR were processed using the MDA 180 analyzer (BioMerieux, Durham, North Carolina). The thromboplastin used for INR measurements was Innovin reagent (Dade Behring Inc, Deerfield, Illinois), which has an International Sensitivity Index of 1.0. During the study, warfarin dose adjustments were based on the plasma INR results.

Complete blood counts were obtained at least weekly and the results of those hematocrit data were each recorded coinciding with the INR samples tested. Anemia was defined as a hematocrit < 32%. Most of the patients were intermittently treated by protocol with an erythropoietic agent and some patients required blood transfusions.

Results

All 39 chronically anticoagulated patients receiving hemodialysis were approached for study participation. The 37 study patients provided 258 blood sample triplicates over the time period of the study. The demographic information for these study participants is provided in Table 1. The mean age of patients participating was 73 yr and nearly half were women. The indications for warfarin anticoagulation were venous thrombosis, pulmonary emboli, chronic atrial fibrillation, antiphospholipid syndrome, artificial (heart) valve, or a combination. There were none taking lower-dose warfarin for dialysis access patency. Seven patients did not fulfill the planned six samples because of medical complications in their dialysis schedules, but all were results were included.

The correlation between POC and plasma INR values was quite high (Figure 1). Of the 258 data pairs, POC INR results were in agreement with Lab INR by ±0.2 INR 67% of the time and ±0.4 INR 89.2% of the time. There was a single occurrence of an observed difference of >1.0 INR (0.38%). Observed differences of 0.7 or more occurred in 12 of the 259 (4.6%), of which seven occurred when the INR was elevated to 4.0 or higher. There is a known low bias for INR results exceeding 5 with the device that we used.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Point-of-care (POC) and plasma International Normalized Ratio (INR) comparisons. (A) Scatterplot of standard plasma laboratory INR (Lab INR) versus POC INR with line of equality for the entire cohort of 258 samples. (B) Bland–Altman plot of the difference of Lab INR − POC INR versus the mean of the two measurements. Horizontal lines represent the mean difference = 0.08 and 95% limits of agreement −0.54 to 0.70. Correlation between Lab INR and POC INR: r = 0.937, P < 0.0001.

To address the second objective regarding the effect of anemia on the absolute difference between Lab INR and POC INR, paired INR measures were compared separately for both anemic and nonanemic patients using the accepted hematocrit cutoff of 32%. Anemia status did not appear to have a clinically important effect on the magnitude of the difference between POC and Lab INR (Figure 2, A and B). Table 2 presents the number of differences for INR of >1, >0.8, and >0.6, comparing both anemic and nonanemic patients. Anemic patients showed a small likelihood to have a slightly lower POC in 60% of the measurements (Figure 2, A and B). This tendency to have a slightly lower POC INR measurement was reflected in the overall t test, which revealed the statistical bias. Despite this bias, the magnitude of the differences shown was small but did not appear important in the vast majority.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Effect of anemia on POC INR measurements. (A) Side-by-side box plots showing the distribution of differences between Lab and POC INR measurements for anemic and nonanemic patients (anemic patients defined as hematocrit < 32%, n = 84; nonanemic defined as hematocrit ≥ 32%, n = 174). (B) Plot of difference of plasma INR − POC INR measured over range of measured hematocrit in all 258 pairs (r = −0.219, P = 0.004).

Our third objective was to compare POC INR measures from blood sampled by finger stick and the patient's dialysis access (central venous catheter, AV fistula, or dialysis graft) each to the standard plasma INR. A strong correlation was noted between INR measures obtained from finger stick and either AV fistula or dialysis graft and plasma INR (Figure 3A). In contrast, blood obtained from a central venous catheter did not correlate well with standard phlebotomy plasma INR values that we knew had heparin contamination (Figure 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Sampling source and POC INR accuracy. (A) Scatterplot of Lab INR versus POC INR samples taken from dialysis arteriovenous (AV) fistula or graft with the line of equality (n = 96). Lab INR graft-obtained POC INR analysis: Mean 0.0573, SD 0.253, 95% confidence interval 0.006 to 0.109. (B) Scatterplot of Lab INR versus POC INR. Both samples taken from dialysis catheter (usually an Ash-split catheter) with the line of equality (n = 139). Catheter-obtained POC INR results showed poor agreement with Lab INR results.

Discussion

The principal finding of this study is that POC INR measures using the Roche CoaguChek S device compare quite favorably with plasma INR values in hemodialysis patients on chronic warfarin therapy. With the availability of new technology that is more convenient, efficient, and potentially cost-effective, the general applicability to particular patient populations requires validation. POC testing provides expedient INR results that enable health care providers to perform prompt warfarin dosing adjustment decision-making and face-to-face education with the patient. This management paradigm improves compliance and provider and patient satisfaction while reducing both bleeding and thrombotic complications (8). Despite complex hematologic and electrolyte derangements associated with ESRD and intermittent dialytic therapy, POC INR testing did not show impairment relative to plasma INR determinations. This finding is anticipated to ease the management of these complicated patients requiring chronic warfarin therapy. Our results showed a high degree of correlation between the POC INR method and standard Lab INR (Figure 1) shown in our data (Table 2) in our 258 paired samples. We searched several INR magnitudes within our POC INR results for any breakdown in agreement with Lab INR and found that 67% of results were within ±0.2 INR and ±0.4 INR 89.2% of the time. The mean difference of 0.08 with a SD of 0.31 met our goal for concordance of POC testing. Similar studies have found acceptable mean differences (±SD) of between 0.2 (±0.31) and 0.8 (±0.68) to establish an INR concordance between POC methods and Lab INR (2). Our data showed 16 instances (6%) in which the difference between Lab INR and POC INR was >0.6, only four (1.5%) when the difference was >0.8, and just one (0.38%) in which the difference exceeded 1.0 INR unit, which could be a data outlier and arguably excludable.

The second major finding of this study is that POC INR-measured results were not necessarily impaired by low hematocrit counts. Most POC INR testing device manufacturers list anemia as a potential limiting factor in INR measurement accuracy with their device (6). Anemia is a common complication in hemodialysis patients, but did not appear to clinically impair our test accuracy. We analyzed our data for discernible differences between the two INR methods specifically in comparing those with anemia versus nonanemic and found them lacking. The 16 paired results showing an INR difference of >0.6 occurred similarly in both anemic and nonanemic patients. Additionally, result differences of >0.8 INR all occurred in patients who were proven not anemic (Table 2). Correlation results remained high in our anemic patients compared with those nonanemic as shown in the plot of the INR difference (Figure 2B). However, there was noted bias in the anemic group for a slightly lower POC INR measurement as depicted in the box plot (Figure 2A) and difference in calculated SD. Figure 2B depicts the slight difference graphically over our entire hematocrit range. Although samples drawn in anemia (n = 83) showed this slight statistical bias, it did not appear to have a clinically significant effect on the magnitude of the difference between POC and Lab INR. In other words, the finding of a small but significant tendency for anemic patients to have a slightly lower POC INR result compared with Lab INR results did not tend to affect the clinical utility of the method of POC INR.

The third major finding of this study is that blood sampling from a finger stick, AV graft, or fistula provide similarly acceptable blood source for accurate INR assessment. In contrast, blood sampling from a central venous catheter did not compare well with plasma INR data, likely because of heparin contamination. Central catheters for dialysis are routinely utilized for practically obtaining laboratory blood samples, but are instilled with a heparin solution to prevent them from clotting. Heparin-exposed blood samples for POC INR are known to invalidate testing with this device. The correlation of POC INR data for blood samples obtained from these heparin-exposed catheters was terrible despite repeated attempts at thorough catheter flushing techniques (Figure 3). In contrast, the blood sample for POC INR obtained from the patients’ dialysis AV fistula or grafts showed excellent correlation to both the POC INR from finger stick and the standard Lab INR measurement. Dialysis fistulas and grafts are not subjected to heparin as the dialysis catheters are to prevent clotting. We confirmed that blood for POC INR sampled from a central dialysis catheter was confirmed invalid, as stated by the device manufacturer.

Even with our good statistical correlation of POC INR methodology, the applicability of POC INR to clinical practice was a priority. For the purposes of defining clinical applicability we established that an INR outside of the goal range ±0.2 would result in a warfarin dosage adjustment (2). For dosing purposes, discordance between the POC INR and Lab INR occurred when one of the two results indicated a dose change and its corresponding pair INR suggested no change (within our defined goal of 2 to 3 ± 0.2 INR units). We counted the absolute number of inappropriate dosing decisions that would have occurred using the POC device in our dialysis practice (Table 3). There were no instances where one INR result indicated a dose alteration in the opposing direction from its paired INR. Agreement between dosing decisions occurred in 212 (82%) instances in which both the POC and Lab INR results suggested no change or a change of warfarin dose in the same direction; that left 46 (17.8%) instances in which either the POC INR or the Lab INR was suggesting a change, but not both. Of note, discrepant dosing decisions were not any more likely to occur in anemic patients than nonanemic. Had we broadened the allowable INR 2- to 3-goal range from ±0.2 to ±0.3, there would be only 18 (7%) instances in which dosing discordance occurred between the POC and Lab INR.

Limitations in our conclusions include the fact that most of our anemic patients were modestly so at hematocrit 28 to 32% with relatively few INR results to compare in patients with more significant anemia. The issue of POC INR reliability in anemia should be examined in a larger study. Another limitation may be that we obtained our paired samples from 37 chronic dialysis patients and some of those contributed more results than others. Lastly, venipuncture sampling for Lab INR would have been the most ideal comparator; however, dialysis patients with viable access object to venipuncture. In our dialysis centers, blood sampling from a dialysis access whenever possible is the routine preference for the comfort of our patients.

In summary, this POC INR method appears to correlate well with standard Lab INR in our dialysis patients taking warfarin. Although we compared only 84 POC INR results, the concern with using the POC INR method in our anemic dialysis patients was not confirmed. In our anemic patients, other than a modest bias, POC INR methodology compared favorably with the standard Lab INR method despite the caution in doing so from the manufacturer of POC devices. We confirmed that blood obtained for INR from a heparin-exposed catheter cannot be utilized on this POC device; however, a drop of whole blood taken from an AV fistula (graft) before introducing heparin and dialysis worked well. Lastly, therapeutic decisions based on POC INR results in our dialysis population appear valid, with few discrepant decisions made, and the proper use of this technology makes anticoagulation management expedient.

Disclosures

None of the authors has any conflicts of interest, sponsorships, or relationships potentially influencing any part of this study.