The model for end-stage liver disease (MELD) is a mathematical score originally developed to predict survival after creation of a transjugular intrahepatic portosystemic shunt in patients with cirrhosis1 and later ostensibly used to prioritize patients for liver transplantation in the United States.2 The MELD score is constructed by including in a mathematical function such variables as creatinine, bilirubin, and prothrombin time (PT), with results expressed as international normalized ratio (INR). The rationale of using the MELD score to assign priority for liver transplantation rests on the assumption that the score would be the same across a given area or country if the methods used to measure the variables are standardized and yield the same numerical results regardless of the testing laboratory. This system would ensure national parity for transplantation and would prioritize patients solely on the basis of objective measures. Trotter et al.3 were among the first to realize that specific laboratory methodologies performed in different centers may considerably influence the numerical value of the MELD score and therefore the priority for liver transplantation. According to these authors, the parameter that most influences the MELD score is the PT-INR.3 Other studies spanning more than 10 years have documented that the PT-INR is in fact not standardized across centers when it is used for patients with liver disease.4–7 However, the extent of variability of the PT-INR in the setting of liver disease that may be expected when testing is performed with a vast array of commercial thromboplastins is poorly defined. Furthermore, scanty information is available on how to resolve this issue. This study was designed to provide information on the variability of results obtained with 7 widely used commercial thromboplastins and to explore suitable alternatives to obviate such variability.
The model for end-stage-liver-disease (MELD) is a mathematical score used to prioritize patients for liver transplantation and includes results for creatinine, bilirubin, and prothrombin time (PT) expressed as international normalized ratio (INR). The rationale of using the MELD rests on the assumption that the score would be the same across the country if the methods used to measure the variables yield the same numerical results regardless of the testing laboratory. Evidence was provided that specific methodologies may influence the MELD, and the PT-INR was identified as the most important. This study was designed to provide information on the between-thromboplastin variability and to explore alternatives to obviate such variability. Fifty-seven patients with cirrhosis were selected, and their PTs were measured with 7 thromboplastins. The thromboplastins were previously calibrated by testing plasmas from patients on vitamin K antagonists and healthy subjects to assign the international sensitivity index (ISIvka) needed to convert PT into INR. Each of the thromboplastins was also assigned an ISIliver by substituting in the calibration the plasmas from vitamin K antagonist patients with plasmas from patients with cirrhosis. INR and MELD values for individual patients were calculated by using the ISIvka or the ISIliver. The mean INRvka obtained with the 7 thromboplastins were significantly different (P < 0.001). Conversely, the mean INRliver were not. Similarly, the mean MELDvka were significantly different (P < 0.001), but those differences were abrogated for the MELDliver. Conclusion: The alternative thromboplastin calibration using plasmas from patients with cirrhosis instead of from vitamin K antagonist patients is feasible and may resolve the variability of the MELD to prioritize patients for transplantation. (HEPATOLOGY 2007.)
Patients and Methods
Fifty-seven adult patients with liver cirrhosis of graded severity according to the Child-Pugh score (Child A, n = 20; B, n = 19; and C, n = 18) were selected among those who attended the clinic for phlebotomy for routine clinical indications. Their PT-ratio (patient-to-normal coagulation time) values as determined with the routine thromboplastin ranged from 1.06 to 2.79. Cirrhosis was diagnosed on the basis of clinical, laboratory, and ultrasonographic evidence. Criteria for exclusion were the use of drugs known to interfere with blood coagulation, extrahepatic malignancy, and known hemostatic disorders other than cirrhosis. The cause of cirrhosis was hepatitis C virus (n = 26), hepatitis B virus (n = 7), alcoholic (n = 16), viral and alcoholic (n = 4), or cryptogenic (n = 4). None of the patients had chronic cholestatic disease, as evidenced by the levels of serum alkaline phosphatase. Sixty patients on vitamin K antagonists were randomly selected among the population of patients who attended the anticoagulation clinic for regular check of the INR and dose prescription. All of them were stabilized (more than 6 months) on vitamin K antagonists within the range of anticoagulation corresponding to 1.5 to 4.5 INR. Twenty healthy subjects recruited from the population of the laboratory staff and medical students were also included in the calibration studies. None of them had family or personal history of bleeding or thrombotic diseases and were not taking drugs known to interfere with hemostasis. After approval of our Institutional Review Board and informed consent of the patients and healthy subjects, blood samples were collected by clean venipuncture in vacuum tubes (Becton Dickinson, Meylan, France) containing 105 mM trisodium citrate as anticoagulant (ratio blood: anticoagulant, 9:1). Blood was centrifuged at 2,000g (room temperature) for 20 minutes, and the supernatant plasma was harvested, aliquoted in capped plastic tubes, frozen with liquid nitrogen, and stored at −70°C until testing.
Seven commercial thromboplastins (see Table 1 for details) were identified among those widely used in North America and Europe, and suitable amounts were purchased from their manufacturers. Stocks of these working thromboplastins were stored as recommended by the manufacturers until testing. One specific lot (N453415) of human recombinant thromboplastin (Recombiplastin, Instrumentation Laboratory, Orangeburg, NY) was used as secondary standard to determine the international sensitivity index (ISI) for the 7 working thromboplastins. This secondary standard has been previously calibrated by 3 reference laboratories against the human recombinant standard rTF/95 held by the World Health Organization (WHO)8 and assigned an ISI of 0.931.
|Recombiplastin||Instrumentation Laboratory||Human recombinant||N958473|
|Thrombotest||Axis Shield||Ox brain combined*||10121290|
|Innovin||Dade Behring||Human recombinant||536950|
|Thromborel S||Dade Behring||Human placenta||505783|
|Neoplastin plus||Stago||Rabbit brain||664768|
|PT HS||Instrumentation Laboratory||Rabbit brain||N353581|
|Simplastin Excel||Biomerieux||Rabbit brain||132400|
Testing with all thromboplastins was performed in combination with a photo optical coagulometer (Electra MLA 1600, Instrumentation Laboratory, Lexington, MA) according to manufacturer specifications. PT results for each subject and thromboplastin were recorded as clotting times (seconds). Other measurements such as bilirubin and creatinine were performed according to standard laboratory techniques.
Conventional Calibration of Working Thromboplastins.
The ISI of the 7 working thromboplastins was determined according to the WHO guidelines.9 Briefly, parallel PT testing (single determinations) with each of the working thromboplastins and the secondary standard was performed for 60 plasmas from patients on vitamin K antagonists and 20 from healthy subjects. The numbers of plasmas to be included in the calibration may be smaller than 60 and 20, provided that the CV of the slope of the orthogonal regression line is 3% or less.9 PT values obtained with each working thromboplastin were plotted on a double-log scale versus those obtained with the secondary standard (working thromboplastin on the horizontal axis), and the best fit line was drawn through the data-points. The slope of the regression line was estimated by orthogonal regression analysis, and this value was multiplied by the ISI of the secondary standard to derive the final ISI for each working thromboplastin. The accuracy and precision of each calibration was judged by looking at the relevant parameters set by WHO (i.e., precision of calibration expressed as the CV of the slope of the orthogonal regression line and the identity of the patients-only and the normals-only regression lines).9
Alternative Calibration of Working Thromboplastins.
The ISI of the 7 working thromboplastins was also determined by plotting on a double log-scale paired PT values for patients with liver cirrhosis (instead of patients stabilized on vitamin K antagonists) and healthy subjects as determined with the working thromboplastins and the secondary standard. The parameters of the regression line were estimated as specified for the conventional calibration.
Calculation of the INR.
PT values for each patient and thromboplastin were converted into INR by using 2 different equations: INRvka = (PTpatient/MNPT)ISIvka or INRliver = (PTpatient/MNPT)ISIliver, where the MNPT (mean normal PT) is the geometric mean PT of the 20 healthy subjects and the ISIvka and the ISIliver are the ISIs determined using plasmas from patients on vitamin K antagonists or from patients with cirrhosis. The INR values for the 57 investigated patients obtained with each thromboplastin were then used to calculate mean values. The dispersion of the resultant 7 mean values was estimated by calculating the overall mean, the SD, and the CV according to the equation: CV = (SD/mean) × 100. The CV value was taken as an index of the between-thromboplastin dispersion of the INR. Accordingly, the higher the CV, the greater the dispersion.
Calculation of the MELD Score.
This was calculated for each patient using the following equation2:
For each patient, the INR value inserted into the equation was the one calculated by the conventional calibration (INRvka) or by the alternative calibration (INRliver). The dispersion of the MELD score results (MELDvka vs. MELDliver) was judged by the CV value of the overall mean as for the INR (see previous discussion). Results are reported as means and SD. The analysis of variance for repeated measurement was used to test for differences of mean values, and a P value of 0.01 or less was considered statistically significant. All statistical analyses were performed with the SPSS software package, version 13.0 (Chicago, IL).
The parameters of thromboplastin calibrations with conventional or alternative methods are shown in Table 2. The CV of the slope, which can be taken as an index of the precision of calibration, was in all instances acceptable (less than 3%), except for the ISIliver of Thrombotest, which probably would have required greater numbers of plasmas for calibration. All thromboplastins except Recombiplastin and Thrombotest displayed ISIvka values greater than the correspondent ISIliver. The greatest difference was observed for the rabbit brain Neoplastin (53.7%) and the smallest for the human recombinant Innovin (11.8%). The ox brain Thrombotest and the human recombinant Recombiplastin recorded the smallest absolute difference (−5.7% and −8.6). The 3 rabbit brain thromboplastins that had the greatest ISIvka (Neoplastin, 1.297; PT HS, 1.463; and Simplastin, 1.768) displayed the greatest percentage difference (53.7%; 42.2%, and 55.0%). The responsiveness of the thromboplastins to the coagulopathy induced by liver cirrhosis relatively to the international standard was the greatest for Neoplastin and Innovin (ISIliver, 0.844, and 0.845) and the smallest for Simplastin (ISIliver, 1.141). All the thromboplastins displayed responsiveness to the coagulopathy induced by liver cirrhosis similar to or better than the international standard. The two thromboplastins that displayed the lowest responsiveness to the defect induced by vitamin K antagonists (Simplastin and PT HS) were also the least responsive to the defect induced by liver cirrhosis.
|Thromboplastin||ISIvka*||CV of the Slope†||ISIliver‡||CV of the Slope†||%Diff.§||MNPT‖ (sec)|
INR for Patients with Liver Cirrhosis
The distribution of individual INR values for patients with cirrhosis as calculated with ISIvka or ISIliver values for each thromboplastin is shown in Figs. 1 and 2 and Table 3. Mean between-thromboplastin differences were statistically significant for the INRvka (P < 0.001), but not for the INRliver (P = 0.07). The smallest mean (SD) INRvka value was recorded for Recombiplastin [1.29 (0.25)] and the greatest for Simplastin [1.57 (0.55)]. The post-hoc multiple comparison test showed that the mean INRvka values recorded for Recombiplastin and Thrombotest were not significantly different (P = 0.52). The smallest and the greatest mean INRliver values were recorded for Innovin [1.30 (0.28)] and for PT HS [1.34 (0.32)], respectively. The between-thromboplastin CV values were 8.5% and 1.0% for the INRvka and INRliver, respectively. The greatest and the smallest between-thromboplastin variability were recorded for Child C (CV = 12.3%) and Child A (CV = 5.3%) and the intermediate for Child B (CV = 7.4%) patients.
|Thromboplastin||Mean INRvka||Statistical Significance||Mean INRliver||Statistical Significance|
|Recombiplastin||1.29 (0.25)||1.32 (0.28)|
|Thrombotest||1.29 (0.26)||1.32 (0.28)|
|Innovin||1.34 (0.33)||1.30 (0.28)|
|Thromborel S||1.44 (0.42)||P <0.001||1.33 (0.29)||NS|
|Neoplastin plus||1.54 (0.47)||1.31 (0.24)|
|PT HS||1.54 (0.59)||1.34 (0.32)|
|Simplastin Excel||1.57 (0.55)||1.32 (0.27)|
MELD Score for Patients with Liver Cirrhosis
The distribution of individual MELD score values for patients with cirrhosis as calculated with INRvka or INRliver values for each thromboplastin is shown in Figs. 3 and 4. Mean between-thromboplastin differences were statistically significant for the MELDvka (P < 0.001) but not for the MELDliver (P = 0.02). The smallest mean (SD) MELDvka value was recorded for Recombiplastin [12.24 (7.31)] and the greatest for Simplastin [14.15 (8.23)]. The mean MELDliver values obtained with all thromboplastins were almost identical and ranged from 12.30 (Innovin) to 12.63 (PT HS) (Fig. 4). The between-thromboplastin CV values were 6.3% and 0.8% for the MELDvka and MELDliver, respectively.
Dependence of the MELD Score from the INR
This was estimated by calculating the percentage difference (individual value from the overall mean value obtained with the 7 thromboplastins) for the MELDvka or the MELDliver and reported as a function of increasing INR. As shown in Figs. 5 and 6, the percentage difference increased with increasing INR for the MELDvka (Fig. 5), but not for the MELDliver (Fig. 6).
The PT test was developed in 1935 as a means to investigate patients with obstructive jaundice10 and later used to monitor patients treated with vitamin K antagonists.11 The between-thromboplastin variability of results, which was observed as a consequence of the coagulation defect induced by vitamin K antagonists, was a cause for concern and fostered research work aimed at finding appropriate solutions to harmonize PT results. These efforts culminated in the 1980s with the development of the new scale of reporting PT values named INR.12 The INR was later endorsed by the WHO13 and represents now the scale of choice to report PT results for patients treated with vitamin K antagonists.9 By means of this system, commercial thromboplastins are calibrated by their respective manufacturers by testing plasmas from patients on vitamin K antagonists and from healthy subjects with the thromboplastin to be calibrated and 1 of the international thromboplastin standards available from WHO. The slope of the line of relationship between paired PT values can be taken as a measure of the responsiveness to the coagulation defect induced by vitamin K antagonists of the working thromboplastin relative to the WHO standard and is called ISI. The ISI can be in turn used to convert PT results into INR. By definition, if all thromboplastins are calibrated against 1 of the WHO standards, their results are interrelated and the between-thromboplastin variability greatly reduced. The model of calibration implies that the INR scale is valid only for patients treated with vitamin K antagonists and that its use in other clinical situations should be carefully evaluated. Unfortunately, this limitation was not fully appreciated, and the INR scale has been widely used without preventive validation to report PT results also for patients with liver disease, especially for the calculation of the MELD score to prioritize patients listed for liver transplantation.2
The results of this study confirm and extend previous observations that patients with liver cirrhosis might achieve MELD scores that are dependent on the thromboplastin used for testing the PT-INR and therefore may achieve variable priority for transplantation. The average difference recorded for the investigated thromboplastins [from the highest (14.15) to the lowest (12.24)] was 1.91 MELD points, but it can be greater if one looks at the individual values (see for instance those patients with MELD score greater than 20 in Fig. 3). In spite of the fact that the between-thromboplastin variability of the INR (and the MELD score) for patients with liver disease has been recognized for many years, few attempts have been made to tackle this problem.5 We thought that a system of calibration similar to that undertaken for patients on vitamin K antagonists also could be feasible for patients with liver disease. The proposed system calls for testing PT for plasmas from patients with liver disease instead of patients treated with vitamin K antagonists and to insert them in the calibration model. The feasibility of the alternative system of calibration was evaluated by using 7 different commercial thromboplastins and proved highly reliable for the following reasons. First, it fulfilled for all but 1 thromboplastin the main requirement for calibration set by WHO,9 which requires that the CV of the slope value as determined by orthogonal regression analysis be equal to or smaller than 3% (Table 2). The alternative calibration also fulfilled another WHO requirement that calls for the orthogonal regression line drawn through the patient's data points to pass through the mean of the data points from healthy subjects (not shown). Second, the alternative calibration was able to harmonize not only the INR results for patients with liver disease (Fig. 2) but also the resultant MELD score (Fig. 4). In both cases, the between-thromboplastin variability expressed as the CV value was greatly reduced (from 8.5%-1.0% for the INR and from 6.7%-0.8% for the MELD score, Figs. 1–4). The MELD score of individual patients calculated with the INRliver was practically the same regardless of the thromboplastin used for testing (see Fig. 4 as opposed to Fig. 3). After adequate validation, this system of calibration could be implemented similarly to that of the INR for patients treated with vitamin K antagonists at a reasonable extra cost. Manufacturers of commercial thromboplastins should provide 2 different ISI values for their thromboplastins: the regular ISI value for patients on vitamin K antagonists and the alternative ISI value valid for patients with liver disease. Alternatively, the ISIliver might be locally determined by using certified calibration plasmas in a manner analogous to that which is being developed for local calibration for patients on vitamin K antagonists.14 However, the source, characteristics, and method of use for these calibration plasmas should be thoroughly investigated. The scientific societies for the study of liver disease could undertake the preparation of the guidelines for calibration in a manner analogous to that used for the preparation of the guidelines issued by WHO for anticoagulated patients.9 The partnership between scientific societies and commercial manufacturers would be highly beneficial for the standardization in this field and ultimately for prioritization of patients awaiting liver transplantation. A role in this system also could be played by the small easy-to-run coagulometers, which are now frequently used as near-patient-devices for PT-INR testing. Numbers of studies15, 16 have shown that these devices fulfill the WHO requirements for conventional thromboplastin calibration. If this also holds true for the alternative calibration, these devices might be helpful for the determination of the MELD score.
The study has the following limitations. First, among the 3 variables that serve as parameters to calculate the MELD score, only the effect of the INR has been evaluated, and it is not known what the additional impact of the variability would be, because of the measurement of bilirubin and creatinine. This also explains why relatively narrow MELD differences have been recorded in this study (1.91 points, on average) as opposed to those obtained by Trotter et al.3 (approximately 4 points), who took into account the other 2 variables. Second, the between-thromboplastin variability as investigated here cannot be necessarily translated to the “real life” because the thromboplastins have been calibrated in 1 laboratory and in combination with 1 coagulometer only. The real variability may be greater when assessed by using different types of coagulometers, and this needs to be investigated in more detail. Third, we enrolled in this study relatively few patients with Child C and very long PT. Therefore, we cannot assess properly the effect of the between-thromboplastin variability on patients with more severe end-stage liver disease listed for liver transplantation. However, the differences in MELD score (Fig. 5) increase with increasing INR. Therefore, the magnitude of the between-thromboplastin variability of the MELD score would be even greater than that found in this study had we investigated patients with more severe end-stage liver disease. Fourth, we did not investigate the possible role of the cholestatic-associated vitamin K deficiency, which presumably may affect the determination of the ISI and the INR. Whether administration of vitamin K before the INRliver determination may be beneficial in this setting remains to be determined. In this study, the same set of plasmas from 57 patients with cirrhosis has been used to determine the ISIliver (calibration plasmas) and to test the hypothesis that the alternative calibration can be used to harmonize results (test plasmas). Different sets of calibration and test plasmas would have been more appropriate, but this would have required a greater number of plasmas. Conversely, splitting the 57 plasmas between the 2 groups would have made the final number in either group too small to draw firm conclusions. Therefore, we elected to carry out a post hoc parsimonious statistical analysis whereby some of the plasmas (i.e., the 22 with MELD score ≥14, who are presumably those of interest in the present context) were removed 1 by 1 from the calibration. This generated 22 different calibrations for each thromboplastin where the calibration plasmas were 56 instead of the original 57. The ISIliver resulting from the removal of each plasma was then used to convert its PT into the INRliver. As expected, the 56-plasma calibrations were only marginally different from their counterparts with 57-plasma calibrations. This parsimonious design allows the 22 calibration plasmas to be used as test plasmas without interference.17 The results of this post hoc analysis did not alter the statistical significance nor the conclusions of the study (not shown).
In conclusion, this study confirms end extends previous observations on the unsuitability of the INR scale devised for patients on vitamin K antagonists to be used for patients with liver disease and to calculate the MELD score to prioritize patients for liver transplantation. The design of the study in which all thromboplastins were calibrated centrally against the same international standard and according to the WHO requirements demonstrates that the different INR values currently recorded for patients with liver disease are genuine and not attributable to inaccuracy of the ISI calibration. The possible mechanisms underlying the differences in ISI for patients with chronic liver disease and for patients on vitamin K antagonists have been reviewed18 and may rest on the variable deficiency of such non–vitamin K–dependent coagulation factors as factor V and fibrinogen, which occurs in patients with cirrhosis but not in patients on vitamin K antagonists. Furthermore, chronic liver diseases do not have the same effect on the relative activities of vitamin K–dependent coagulation factors as treatment with vitamin K antagonists.19 The observation that the alternative calibration proposed here (use of plasmas from patients with liver disease instead of plasmas from patients on vitamin K antagonists) was able to harmonize results of the INR stands in favor of this concept and suggests that the alternative harmonization system could be considered for implementation, provided that further studies including more patients, thromboplastins, and coagulometers confirm the results. Furthermore, clinical studies should be carried out to assess the role of the new INRliver in the MELD score.