SEARCH

SEARCH BY CITATION

Keywords:

  • Liver allocation;
  • liver transplantation MELD

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References

Priority for liver transplantation is based on the Model for Endstage Liver Disease (MELD) score, a mathematical function which includes international normalized ratio (INR). We present an analysis to determine the lab-to-lab variation in INR at 14 clinical laboratories across the United States. We performed a survey to identify representative clinical laboratories across the United States, where INR was measured in the determination of MELD score. Five ‘standard’ samples for INR were formulated and were sent to the 14 clinical laboratories to determine variation in INR and MELD score. Among the 14 clinical laboratories, the range in INR for the five samples was: sample 1 (1.2–2.0), sample 2 (1.4–2.5), sample 3 (1.7–3.4), sample 4 (1.9–3.7) and sample 5 (2.4–5.1). The range in calculated MELD score was: sample 1 (8–14), sample 2 (10–17), sample 3 (12–20), sample 4 (14–21) and sample 5 (16–25). The selection of the clinical laboratory used to determine INR may result in substantial changes in MELD score independent of severity-of-illness. These data suggest that further review of interlaboratory variation in MELD should be undertaken because of the potential impact on prioritization for liver transplantation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References

Approximately 10 years ago, a rancorous debate emerged regarding the allocation of deceased donor (DD) livers in the United States. At that time, the basis of liver allocation was a scoring system whose major determinants included waiting time as well as subjective assessments of severity-of-illness. While this system had served the liver transplantation community well for several years, it was devised when the number of patients listed for liver transplantation was relatively small. As the national transplant list grew by 10-fold between 1990 and 1999, competition for the limited supply of organs intensified. Consequently, physicians utilized calculated measures to increase the priority for patients with the greatest perceived need for transplantation. The most common means to increase rank on the transplant list was to enhance the subjective assessment of severity-of-illness. As such practices became commonplace, the nationwide inequities in liver allocation were increasingly apparent leading to a public perception that the allocation system had become corrupted. As a result, the Institute of Medicine (IOM) performed a comprehensive review of the liver allocation system in 1999 (1) and concluded that (1) a large disparity of the nationwide access to transplant existed, (2) liver allocation should be based on an objective scoring system predictive of need for transplantation and (3) waiting time is irrelevant to the need for transplantation and should be eliminated from the allocation scheme. Following this assessment, the Model for Endstage Liver Disease (MELD), a mathematical function predictive of 3-month mortality, was developed and implemented nationwide as the basis for liver allocation in February 2002 (2,3). The MELD score is calculated by the following equation: MELD score = 10(0.957ln(serum creatinine, mg/dL) + 0.378ln(bilirubin, mg/dL) + 1.12ln(INR) + 0.643), where laboratory values <1.0 are set to 1.0 for purposes of the MELD score calculation and the maximum serum creatinine considered within the MELD score equation is 4.0 mg/dL. The INR has the highest multiplicative value of the three variables. The MELD score fulfilled the provisions of the IOM review in that prioritization for transplantation was based solely on objective variables (total bilirubin, creatinine (Cr) and international normalized ratio (INR)) and waiting time was essentially eliminated as a determinant. In addition, the MELD allocation scheme was designed to eliminate subjective ‘upgrading’ on the transplantation list. As a result, allocation of livers under the MELD system was intended to achieve nationwide parity in access to liver transplantation, as reflected in the following statement by UNOS: ‘the MELD and PELD formulas are simple, objective and verifiable and yield consistent results whenever the score is calculated’.

However, in a preliminary analysis, we have found significant variability in one of the determinants of the MELD score, the INR, based on the selection of the laboratory methods used for its determination (4). INR is defined by the following equation, INR = (patient's prothrombin time/mean normal prothrombin time)x, where x = International Sensitivity Index. A study from our center showed that blood samples from the same 29 patients analyzed at three different clinical laboratories resulted in a 26% variation in INR translating to a significant 20% difference in MELD score. As an extension of these initial findings, we performed the current analysis in representative clinical laboratories across the United States to determine (1) the interlaboratory variation in INR over the range of INR values found in patients listed for liver transplantation and (2) if the differences in INR translate into clinically relevant changes in MELD score and therefore priority for transplantation.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References

The two phases of this study were approved by the Colorado Multiple Institution Review Board (COMIRB) and were devised with input from members of the Coagulation Resource Committee of the College of American Pathologists (JO, JL). First, we identified the clinical laboratories where INR was measured for the calculation of the MELD score in 139 patients listed for liver transplantation at two geographically diverse transplant centers (the University of Colorado Health Sciences Center and Duke University Medical Center). We recorded the city, state and type of facility (community vs. transplant center) for each clinical laboratory. Then we surveyed these clinical laboratories for the following information: type of INR measuring device, thromboplastin reagent and international sensitivity index (ISI). We selected 14 representative clinical laboratories for the next phase of the study for which we collected blood samples from patients listed for liver transplantation at the University of Colorado Health Sciences Center, as shown in Figure 1. All blood for coagulation testing was drawn into a ‘purple-top’ tube (containing 3.2% buffered sodium citrate v/v: 9 parts whole blood/1 part citrate). After the INR was measured for clinical indications, an aliquot of 113 samples was saved and frozen at −80°C for less than 3 months. The 113 aliquots were then ranked by INR value from lowest to highest. To obtain ‘standard’ samples across the spectrum of INR values, all of the aliquots between the following INR values were combined to form five ‘standard’ samples: #1 – 0.9 to 1.5, #2 − 1.51 to 2.0, #3 − 2.01 to 2.5, #4 − 2.51 to 3.0 and #5 − >3.0, as shown in Figure 1. Finally, the INR of each of the five ‘standard’ samples was measured at our center's clinical laboratory and then divided into 13 aliquots, refrozen and shipped on dry ice to the 13 other outside clinical laboratories. The 13 other clinical laboratories (beside the University of Colorado Health Sciences Center) include: University of Nebraska Medical Center, University of California at Los Angeles, University of Washington (Seattle), University of Texas-San Antonio, Northwestern Memorial Hospital (Chicago), Duke University Medical Center (Durham, NC), Kaiser Permanente Regional Reference Lab (Denver), Hanover Medical Specialists (Wilmington, NC), Piedmont HealthCare (Statesville, NC), Penrose Hospital (Colorado Springs, CO), Quest Diagnostics Incorporated (Denver), Laboratory Corporation of America (Englewood, CO), NorthEast Medical Center (Concord, NC).

image

Figure 1. Creation of five ‘standard’ INR samples. INR = international normalized ratio; *= as measured in the University of Colorado Health Sciences Center clinical laboratory.

Download figure to PowerPoint

Because we had previously noted a trend towards greater interlaboratory variation in INR with higher INR values, we formally assessed the variation in INR and MELD of the five ‘standard’ samples between the 14 clinical laboratories using a modification of the Brown-Forsythe test with terms for lab and sample (5). MELD score was calculated with the following assumptions, serum creatinine = 1.0 mg/dL and total serum bilirubin = 1.0 mg/dL. The following conventions utilized by the United Network for Organ Sharing (UNOS) were used in the calculation of MELD score: INR was rounded to the nearest tenth unit and MELD score was rounded to the nearest whole number.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References

From our survey of clinical laboratories, we found that INR was measured at one of the two transplant centers (Duke University or University of Colorado) in 68/139 (49%) and the remainder were assayed at one of 49 community-based clinical laboratories. From the surveys sent to these 51 clinical laboratories, we received 28 responses relative to the type of INR measuring device, thromboplastin reagent and international sensitivity index. To choose the clinical laboratories for the next phase of the analysis, clinical laboratories were selected to reflect the distribution (of location) of clinical laboratories determined from our survey, such that 50% of samples would be measured at community laboratories and the remainder at transplant centers. Therefore, we selected 14 laboratories (seven representative community laboratories and the two transplant center laboratories and identified five additional laboratories) at five geographically diverse transplant centers. (The five additional transplant center laboratories were selected to achieve a balance of 50% community laboratory and 50% transplant laboratory, the distribution determined in our survey.) (See Appendix.)

The INR and calculated MELD score for the five ‘standard’ samples (1–5) at each of the 14 clinical laboratories are depicted in Tables 1 and 2, respectively. The range in INR for the five samples was: sample 1 (1.2–2.0), sample 2 (1.4–2.5), sample 3 (1.7–3.4), sample 4 (1.9–3.7) and sample 5 (2.4–5.1). The range in calculated MELD score for the five samples was: sample 1 (8–14), sample 2 (10–17), sample 3 (12–20), sample 4 (14–21) and sample 5 (16–25) Variability of INR increased as the mean INR increased (p = 0.0174). However, there was no difference in the variability of MELD among the ‘standard’ samples (p = 0.3203).

Table 1.  Distribution of INR values by clinical laboratory
Clinical laboratorySample 1Sample 2Sample 3Sample 4Sample 5
  1. INR = international normalized ratio; TC = transplant center clinical laboratory; SD = standard deviation; C = community clinical laboratory.

TC – 11.61.82.22.53.0
TC – 21.61.82.72.83.5
TC – 31.51.72.32.53.1
TC – 41.91.61.82.12.7
TC – 51.41.51.92.12.7
TC - 61.41.62.12.53.0
TC - 71.51.82.32.83.3
C - 11.31.41.71.92.4
C - 21.21.41.82.02.6
C - 31.41.72.22.53.1
C - 42.02.53.43.75.1
C - 51.61.72.12.53.0
C - 61.71.92.23.0
C - 71.31.51.92.12.7
mean 1.52 1.68 2.15 2.42 3.07
SD 0.23 0.28 0.46 0.47 0.67
Range1.2–2.01.4–2.51.7–3.41.9–3.72.4–5.1
Table 2.  Distribution of MELD scores by clinical laboratory
Clinical laboratorySample 1Sample 2Sample 3Sample 4Sample 5
  1. MELD = Model for Endstage Liver Disease score; SD = standard deviation; TC = transplant center clinical laboratory; C = community clinical laboratory.

TC - 11213151719
TC - 21213181820
TC - 31112161719
TC - 41412131518
TC - 51011141518
TC - 61012151719
TC - 71113161820
C - 1 910121416
C - 2 810131417
C - 31012151719
C - 41417202125
C - 51212151719
C - 612141519
C - 7911141518
Mean10.912.114.916.318.9
SD1.91.82.11.92.1
Range8–1410–1712–2014–2116–25

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References

These results demonstrate a substantial variation in INR and MELD score between representative clinical laboratories across the United States. Consequently, priority for transplantation may be significantly altered based on the selection of the clinical laboratory where INR is measured in the calculation of the MELD score. These findings are contrary to assumptions made by UNOS that the MELD score is ‘simple, objective and verifiable and yields consistent results whenever the score is calculated’ and have important implications for patients on the liver transplantation list (6).

For patients with a high position on the transplant list, small changes in MELD score and priority for transplantation (either higher or lower) may represent the difference between an imminent transplant and an indefinite wait. Because these patients have the highest mortality rate on the transplant list, prolongation of waiting time could result in death or removal from the list (due to disease progression) prior to transplantation. For patients with a lower MELD score (≤18), a change in several MELD points could determine eligibility for allocation of a DD liver altogether, because a recent change in UNOS allocation policy dictates that patients with MELD score <15 are not eligible for allocation of a DD graft unless the liver has been rejected for all listed patients in the region with higher MELD scores (7). In addition, the maximum changes in MELD score that we noted were striking. Identical blood samples assayed at different laboratories led to differences in MELD score of up to 9 points with the greatest differences noted at one community laboratory. However, as shown in Table 2, differences in MELD score for identical blood samples were as high as 7 points.

Clinicians may be surprised by our findings, because the term ‘INR’ implies that INR values are ‘normalized’ between clinical laboratories. The INR was devised to standardize the antithrombotic effects of warfarin and improve the clinical management of patients receiving this drug. Standardization of therapeutic anticoagulation became necessary when clinicians noted that the various thromboplastin reagents used to determine prothrombin time differed markedly in their responsiveness to the anticoagulant effects of warfarin (8). To rectify this problem, the World Health Organization (WHO) proposed the INR, defined as INR = (patient's prothrombin time/mean normal prothrombin time)x, where x = International Sensitivity Index. The INR is a correction formula that adjusts for the variable sensitivities of different thromboplastin reagents and allows standardization of prothrombin time ratio (determined by any thromboplastin reagent) to a reference WHO thromboplastin standard (9). With this standardization, the INR improves the safe and effective dosing of warfarin independent of the sensitivity of the thromboplastin reagent. However, the INR was not designed for its use in the MELD score; to provide a reproducibly precise and objective assessment of severity-of-illness in patients with MELD. In fact, numerous studies have demonstrated that the test characteristics of INR in liver patients would poorly serve this function. Furthermore, UNOS has never performed a ‘quality control’ test to determine the variation in MELD score between clinical laboratories based on the interlaboratory variation in INR.

Robert and Chazouilleres documented the mean INR and range of INR for 29 patients on anticoagulants and in 27 patients with liver disease for seven types of thromboplastin reagent (10). They found a large and statistically significant variation in the mean INR for the 27 liver patients among seven specific thromboplastins. The range of the mean INR for the thromboplastin reagents was 2.3 to 4.1 (40% difference) for the patients with liver disease, p = 0.007. The difference in INR was greatest in patients with the highest INR values, varying by as much as threefold. As expected, the difference in INR for the anticoagulation patients was much smaller, range 3.0 to 3.6 (16% difference), p = ns. Three other studies have reported similar results (4,11,12). Another study using three different thromboplastins found similar variation in the mean INR in 29 liver patients, but no difference in 31 patients on warfarin (11). The range in mean INR for the liver patients was significantly different: from 1.88 with recombinant human thromboplastin to 2.63 with rabbit brain thromboplastin which is a 29% difference, p < 0.0001. However, the range in INR for the anticoagulation patients was not significantly different at 2.63 to 2.75 which is a 4% difference, p = ns. Denson et al. compared INR values in 20 patients with liver impairment using human and rabbit thromboplastin and reported a difference in INR between the two reagents of 25% (12). As noted above, an earlier study from our center (see Introduction) replicated these findings in patients listed for liver transplantation (4).

The reason(s) for the wide variation in INR among liver patients compared to patients receiving warfarin is not known. One explanation may be the different mechanisms of prothrombin time elevation caused by warfarin and liver disease. Warfarin causes prolonged prothrombin time through inhibition of the vitamin K-dependent gamma-carboxylation of coagulation factors II, VII, IX and X. In liver disease, the elevated prothrombin time is due in large part to decreased production of all coagulation factors (13). Also, in contrast to patients taking oral anticoagulants, patients with advanced liver disease also have varying degrees of reduced natural coagulation inhibitor synthesis, acceleration of fibrinolysis with the increase fibrin(ogen) degradation products, reduced and abnormal fibrinogen and reduced absorption of vitamin K, all of which may have variable impact on the INR.

We chose to analyze differences in INR and MELD scores across a wide range of INR values, because we had previously noted a trend towards greater interlaboratory variation in INR with higher INR values (4,10). In addition, we formally assessed the variation in INR and MELD of the five ‘standard’ samples between the 14 clinical laboratories. While the INR variability between the 14 laboratories was significant greater with higher INR values, there was no difference in the corresponding MELD scores.

There are four studies evaluating the effect of freezing on prothrombin time, INR and clotting factor stability. The coefficient of variation for five deep-frozen pooled plasmas in anticoagulated patients was only 2.6%–3.3% (14). Another study evaluated the effect of freezing blood samples in 84 patients on anticoagulation therapy at −40°C for less than one month (15). These investigators reported no difference in INR between freshly assayed or frozen samples. Woodhams et al. evaluated the stability of clotting factors and prothrombin time in six normal volunteers whose blood samples were frozen for two months at both −24°C and −74°C (16). They reported <5% difference in the prothrombin times for samples stored at either temperature. Finally, Iazbik et al. measured prothrombin time in nine plasma samples frozen for one week (17). There was no difference in prothrombin time in the frozen versus freshly assayed samples. Therefore, we do not believe that sample freezing could be responsible for the differences in INR values between the clinical laboratories in this study.

What is the appropriate response to our findings which suggest that the MELD score may not yield consistent, reproducible results? The answer may be different for individual patients compared to the representatives of the organ allocation system. Priority for transplantation is of great concern for patients listed for liver transplantation. Most patients on the transplant list routinely track changes in their MELD score, because of their inherent need to achieve the highest possible priority for transplantation. Our results suggest that patients listed for transplantation could potentially increase their position on the transplant list by utilizing a clinical laboratory which yields the highest possible INR value. Such practices could, however, potentially undermine one of the fundamental purposes of the MELD system which is parity in access to transplantation and allocation of DD livers based on an objective score free of subjective factors. For the stewards of the organ allocation system (UNOS), the goals are somewhat different; namely, the fair and equitable allocation of DD organs across the country. To minimize difference in MELD based on INR variation, UNOS could mandate that all clinical laboratories use the same measuring device, thromboplastin reagent and ISI whenever INR is measured for the determination of MELD. However, there are important practical issues to consider. To be effective, the MELD-based liver allocation system must use input variables that are easily and rapidly obtained, i.e. common laboratory tests available at all clinical laboratories across the country. This is due in large part to the widely dispersed and large number of listed patients (more than 10,000), who utilize thousands of different clinical laboratories and may require recalculation of their MELD score as often as once weekly. Our data revealed a wide disparity in the methodologies used in the measurement of INR. Consequently, the likelihood of achieving widespread uniform methodologies for INR determination seems prohibitively difficult. Perhaps the most appropriate response to our findings would be to educate patients and physicians regarding the variation in MELD relative to INR. Thereafter, UNOS may wish to consider a more comprehensive evaluation on the impact of inter-laboratory variation of INR, and perhaps Cr and bilirubin, on MELD score. More important would be to determine if interlaboratory variation has a clinically relevant impact on prioritization for liver transplantation. If clinically relevant differences in prioritization were noted, then UNOS could consider corrective measures to improve parity towards access to DD livers. However, whatever means are used to assess this problem, one of the most important conclusions from our study is that the characteristics of the INR are not well suited for the determination of the MELD score whose goal is equitable nationwide distribution of livers. The wide interlaboratory variation in INR renders this laboratory test inherently problematic for its use in the determination of the MELD score, one of whose primary goals is to ‘yield consistent results whenever the score is calculated’.

In summary, we have demonstrated a clinically relevant interlaboratory variation in INR, and therefore MELD score, which may result in significant changes in prioritization for liver transplantation. These identified variations in INR and MELD score may undermine the fundamental purpose of the current allocation system which is to prioritize patients solely on objective measures. The implications of these findings for patients and the organ allocation system are discussed.

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References
Table Appendix:.  Clinical laboratory, location, device and reagent
LaboratoryDeviceReagentISI
  1. TC = transplant center clinical laboratory; C = community clinical laboratory.

TC – 1Stago STA-Rneoplastin CI+1.3 
TC – 2Trinity Biotech Amax Destinythrombomax HS w/CA1.54
TC – 3Stago STA-Rneoplastin CI+1.23
TC – 4bioMerieux MDA 180 IIbiomerieux PT reagent1.23
TC – 5Sysmex CA-7000Dade Innovin1.02
TC – 6Stago STA-RStago PT reagent1.3 
TC – 7Stago STA compactneoplastin CI+1.27
C – 1Sysmex CA 1500Dade Innovin1.06
C – 2Stago Evolution STA-RDade Innovin0.91
C – 3Stago STA compactNeoplastin CI+1.26
C – 4IL-ACL 8000IL hemosil recombinant1.09
C – 5Stago STA-RNeoplastin CI+1.29
C – 6Sysmex CA-1000Dade Thromboplastin C+1.83
C – 7Dade CA-7000Dade Innovin1.1 

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Appendix
  8. References
  • 1
    Institute of Medicine Committee on Organ Procurement and Transplantation Policy. Organ Procurement and Transplantation: assessing current policies and the potential impact of the DHHS final rule. Washington , DC : National Academy Press; 1999: 129.
  • 2
    Wiesner R, Edwards E, Freeman R et al. Model for end-stage liver disease (MELD) and allocation of donor livers. Gastroenterology 2003; 124: 9196.
  • 3
    MELD score = 10(0.957ln(Cr,mg/dl) + 0.378ln(bilirubin, mg/dl) + 1.12ln(INR) + 0.643), where laboratory values less than 1.0 are set to 1.0 for purposes of the MELD score calculation and the maximum serum creatinine considered within the MELD score equation is 4.0 mg/dl.
  • 4
    Trotter JF, Brimhall B, Arjal R, Phillips C. Specific laboratory methodologies achieve higher model for endstage liver disease (MELD) scores for patients listed for liver transplantation. Liver Transpl 2004; 9951000.
  • 5
    Brown MB, Forsythe AB. Robust Tests for Equality of Variances. J Am Stat Assn 1974; 69: 364367.
  • 6
    United Network for Organ Sharing, http://www.unos.org/SharedContentDocuments/Revised_MELDPELD_2003(2).pdf (accessed December 20, 2005).
  • 7
    Merion RM, Schaubel DE, Dykstra DM, Freeman RB, Port FK, Wolfe RA. The survival benefit of liver transplantation. Am J Transplant 2005; 5: 30713.
  • 8
    Hirsh J, Poller L. The international normalized ratio. Arch Intern Med 1994; 154: 28288.
  • 9
    WHO expert committee on biological standardization. 33rd report. Geneva , Switzerland : World Health Organization: 1983. Technical Report Series No. 687.
  • 10
    Robert A, Chazouilleres O. Prothrombin time in liver failure: time, ratio, activity percentage, or international normalized ratio. Hepatology 1996; 24: 13921394.
  • 11
    Kovacs MH, Wong A, MacKinnon K et al. Assessment of the validity of the INR system for patients with liver impairment. Thrombosis and Hemostasis 1994; 71: 727730.
  • 12
    Denson KWE, Reed SV, Haddon ME, Woodhams B, Brucato C, Ruiz J. Comparative Studies of Rabbit and Human Recombinant Tissue Factor Reagents. Thromb Res 1999; 94: 255261.
  • 13
    Hedner U, Erhardtsen E. Hemostatic disorders in liver disease. In: SchiffER, SorrellMF, MaddreyWC, eds. Diseases of the Liver. Philadelphia : Lippincott Williams and Wilkins, 2003: 625635.
  • 14
    Van Den Besselaar AM, Houbouyan-Reveillard LL, Aillaud MF et al. Multicenter evaluation of lyophilized and deep-frozen plasmas for assignment of the International Normalized Ratio. Thromb Haemost 1999; 82: 14511455.
  • 15
    Grau E, Tenias JM, Olaso MA et al. Monitoring oral anticoagulant treatment from plasma stored for up to 48 hours and frozen plasma. Haematologica 1999; 84: 633636.
  • 16
    Woodhams B, Girardot O, Blanco MJ, Colesse G, Gourmelin Y. Stability of coagulation proteins in frozen plasma. Blood Coagul Fibrinolysis. 2001; 12: 229
  • 17
    Iazbik C, Couto CG, Gray TL, Kociba G. Effect of storage conditions on hemostatic parameters of canine plasma obtained for transfusion. Am J Vet Res 2001; 62: 734735.