Dr A. Tripodi, Angelo Bianchi Bonomi Hemophilia and Thrombosis Center, IRCCS Maggiore Hospital, Mangiagalli and Regina Elena Foundation, University of Milan, Milan, Italy. E-mail: firstname.lastname@example.org
Background Prothrombin time (PT)-derived international normalized ratio (INR) is used to assess bleeding risk and prognosis in cirrhosis, and to guide management of associated coagulation disturbances. Recent studies cast doubt on the validity of the assumptions that form the basis for these applications.
Aims To review and critique the use of the PT-INR in cirrhosis.
Methods Search of the literature.
Results In cirrhosis, there is a decrease in both pro- and anti-coagulants. The PT-INR measures only the activity of procoagulants and fails to capture changes in anticoagulants. It is therefore not surprising that the PT does not predict the bleeding risk. The PT-INR provides a robust measure of liver function but recent data showed INR inter-laboratory variability in this setting. This is not surprising as the INR was validated to normalize results for patients on vitamin-K antagonists, not for cirrhosis. This limitation was not appreciated, but the INR is used to construct the model for end-stage liver disease score to prioritize patients for liver transplantation. Reports showed that model for end-stage liver disease is modified by the thromboplastin used for testing.
Conclusions Alternate tests to predict bleeding risk should be developed. The potential for misuse of the PT-INR should drive the development of alternate algorithms for organ allocation.
In spite of extensive contrary data and in spite of the unique history of the prothrombin time (PT) and the PT-derived international normalized ratio (INR), this test is utilized in the practice of hepatology as a guide to bleeding risk assessment and to procoagulant treatment efficacy. Less controversial, but still hampered by issues of inter-laboratory variation is the use of the PT-INR in measuring liver disease prognosis. In this article, we review and critique the use of this test in liver disease patients and propose specific areas in need of further clinical and laboratory research.
History of the PT and the INR scale
The PT test, defined as the coagulation time (seconds) of a mixture of platelet-poor plasma, tissue factor (thromboplastin) and calcium chloride was developed by A.J. Quick in 1935 as a means to investigate the coagulopathy associated with obstructive jaundice.1 Being responsive to the deficiencies of many coagulation factors (VII, X, V, II and fibrinogen), PT became over the years the test of choice to investigate congenital or acquired coagulopathies as well as to monitor the treatment with vitamin K antagonists (VKA). The key component for the PT test is thromboplastin which may be extractive or recombinant. The variable reagent composition makes coagulation times dependent on the reagent used for testing, thus making inter-laboratory comparability of results difficult. However, this drawback was not considered detrimental to the use of the PT at least until the treatment with VKA became widely used.2 Restriction of patients to move from one laboratory to another and the difficult adoption of the therapeutic intervals (i.e., optimal level of anticoagulation to prevent thrombosis, while minimizing the haemorrhagic risk) became important limitations. Early attempts at expressing PT results as either percentage activity, or the PT-ratio (patient-to-normal coagulation time) failed to harmonize results which were still dependent on the thromboplastin/coagulometer used for testing. At the beginning of the 1980’s a system of calibration was devised3 and endorsed by the World Health Organization (WHO).4, 5 Accordingly, commercial PT systems (defined as the combination of thromboplastin/coagulometer) are now calibrated by their manufacturers against one of the international standards for thromboplastin held by WHO. The system calls for the measurement of PT for plasmas from healthy subjects and patients stabilized on VKA. Paired values obtained with the system to be calibrated (working PT) and with the international standard are then plotted on a log-scale and the best-fit orthogonal regression line is drawn through the data points (see Figure 1). The slope of the line, called international sensitivity index (ISI), is a measure of the responsiveness of the system to be calibrated relatively to the international standard (Figure 1). The ISI can in turn be used to convert PT results obtained with any calibrated working system into the international scale called INR according to the following equation.
The INR is the PT-ratio that would have been obtained had the patient’s plasma been tested with the international standard instead of the working system.5 This concept of calibration leads to important considerations6 which are summarized in Table 1. Most relevant of these considerations to the present discussion is that the ISI and the conversion of PT results into the INR scale are valid only for patients on VKA. This is because PTs from patients on VKA are inserted in the calibration plot (see Figure 1). The logical consequence stemming from this consideration is that the application of the INR scale to the PT measured for conditions other than oral anticoagulant treatment is neither justified, nor secures harmonization of results. Unfortunately, the importance of this limitation has not been fully appreciated and the INR scale is used to construct Child7 and model for end-stage liver disease (MELD)8 scores for patients with liver diseases. This may bring important and unpredictable consequences (see other sections below).
Table 1. Considerations that apply to the international normalized ratio (INR) scale in expressing prothrombin time (PT) results
1. International sensitivity index (ISI) values for working PT measuring systems are interrelated because they are determined against the same World Health Organization (WHO) international standard. As a consequence the INR measured with calibrated working systems minimizes system-related differences.
2. An ISI close to unity means that the working PT systems have the same responsiveness to the defect induced by vitamin K antagonists (VKA) as the international standard, whereas an ISI higher than one means a lower responsiveness.
3. It is important to realize that the calibration model is an approximate one and is intended only to minimize system-related differences, not to abolish them.
4. The INR scale is most accurate in the range 1.5–4.5. This is because only patients on VKA within the above limits of anticoagulation are included in the calibration model (see Fig. 1).
5. The ISI and the conversion of PT results into the INR scale are valid only for patients on VKA. This is because PTs from patients on VKA are inserted in the calibration plot (see Fig. 1).
The PT in liver disease. What do we expect and what does it deliver?
The PT is commonly monitored in patients with liver dysfunction because it is inexpensive, readily available and is commonly perceived to reflect the risk of bleeding. The extent of prolongation of the PT depends on the levels of factors which are synthesized in the liver and are reduced as liver synthetic capacity declines.9 In addition to the pro-coagulant factors, anticoagulant factors such as antithrombin, protein C and protein S are synthesized by the liver and their levels are decreased in liver failure9 (Figure 2). Procoagulant proteins are normally present in excess and deficiencies must be fairly severe to produce haemorrhage.10 By contrast, a more modest reduction in anticoagulant proteins can severely impair the antithrombotic potential of plasma.11 Thus, in a patient with liver failure the defect in the levels of procoagulant factors may be more than balanced by a defect in the levels of anticoagulant proteins.12 Indeed, early reports13 and observations stemming from clinical practice show that many patients with liver disease appear to achieve a balance between reduced pro- and anticoagulant levels and have neither haemorrhage nor thrombosis, until the delicate balance is upset by some other factors such as renal failure14 or infection.15 Recently, evidence has been provided that tissue factor-induced thrombin generation is normal in patients with cirrhosis (Figure 3) even when there are abnormal conventional coagulation tests such as the PT.16, 17 This apparent paradox may be explained if one considers that in cirrhosis the reduction of the procoagulant drive is balanced by the reduction of the anticoagulant drive, thus leaving the balance unaltered.16, 17 Why is this balance not reflected by the conventional coagulation tests? To answer this question, we should consider that one of the major anticoagulant mechanisms operating in vivo (i.e., protein C) needs to be fully activated in order to down-regulate thrombin generation through the selective inactivation of the factors Va and VIIIa.18 The protein C activator in vivo is the complex thrombin-thrombomodulin.18 Thrombomodulin is located on endothelial cells lining the lumen of the vessel wall and much less in plasma.19 The logical consequence is that conventional coagulation tests which operate in plasma where thrombomodulin is virtually absent cannot be fully responsive to the anticoagulant drive. On the contrary, their responsiveness to the procoagulant drive remains intact. Tests for thrombin generation modified by adding soluble thrombomodulin may therefore be more suitable than conventional tests to reflect the coagulation balance which operates in plasmas from cirrhotics.16, 17 These observations lead to the conclusion that the PT may tell us whether a given patient is deficient in procoagulant proteins, but not whether that deficiency is balanced by a deficiency in anticoagulant proteins and this may be one reason that the PT is not good at predicting bleeding risk in patients with liver failure (see below).
This conclusion may also apply to other conditions where there is an acquired deficiency of both the procoagulant and anticoagulant factors, but not to treatment with VKA. At first blush it would appear that liver failure and anticoagulation with VKA would have quite similar effects on the coagulation process. However, this is not true. Most of the coagulation factors depressed by VKA are also reduced in liver disease, but liver dysfunction has effects on coagulation proteins that are not the same as those induced by VKA. Liver failure does not have the same effect on the relative levels of vitamin K-dependent factors as does anticoagulation with VKA. At any given PT-INR, patients with liver failure tend to have lower factor VII levels than anticoagulated patients with the same INR.20 In addition, factor V and fibrinogen levels are reduced in liver disease, but are not affected by VKA. Most importantly, the anticoagulant protein antithrombin is decreased in liver failure along with procoagulant factors, but it is not reduced in patients treated with VKA. These may be the reasons why the PT is good at predicting the risk of bleeding in patients treated with VKA, but not in patients with liver failure. Indeed, the PT was not found to be a predictor of the severity of upper GI bleeding21 in patients with liver disease and studies spanning more than 20 years document that the risk of bleeding following liver biopsy and other invasive procedures is poorly predicted by the PT.22–24
The PT as a measure of bleeding risk
The PT and the PT-derived INR have become established tools in the day-to-day management of liver disease patients especially as indicators of bleeding risk. This has led to a remarkable situation in which the numerical value of the PT and/or the INR is often directly translated into bleeding risk. However, this relationship is remote at best24–26 and perhaps only actual in conditions of super-imposed vitamin K deficiency. At worse, the value of the PT bears almost no relationship to bleeding risk.22, 27–29 However, in spite of the lack of supportive data there is an entrenched practice often based on published reviews or institutional guidelines which utilizes specific numbers as targets of correction for PT in the liver patient failure30, 31 although variation in these cutoffs32 indicates the lack of a sound basis.
The complexity of the bleeding diathesis in liver disease patients, affected by both pro- and anti-coagulant processes16, 17, 33–35 is such that the PT is not capable of providing the sort of information attributed to the test. Moreover, the net bleeding diathesis in these volatile and unstable patients is likely to vary day-to-day or even hour-to-hour depending on the effects of superimposed infection and renal failure which are common and sometimes transient or fatal developments.14, 15 Nevertheless, common clinical practice often includes the use of procoagulants as prophylactic or interventional measures in the face of elevated PT-INR. Controlled clinical trials showed that although such procoagulant agents as recombinant activated factor VII are able to shorten the prolonged PT their clinical efficacy is rather poor.36–38 These findings question the usefulness of the PT in monitoring the response to the therapy with pro-coagulant agents. From a recent informal survey, 72% of respondents either agreed (66%) or strongly agreed (6%) that legal concerns play a significant role in their decision making regarding management of prolonged PT and bleeding risk in liver disease patients.33 What is missing and often not considered is the risk of the intervention in this setting. Furthermore, the target of intervention (correction of the PT-INR) lacks correlation to bleeding risk and, in the setting of portal hypertension, the intervention may actually exacerbate the risk of bleeding if portal pressure is increased.39, 40 This is particularly true of conventional therapy with plasma which carries additional and often unrecognized risks such as transfusion related acute lung injury (TRALI).41
How can these issues be resolved? In the face of the tendency of liver disease patients to suffer spontaneous- or procedure-related bleeding, the first step towards resolving the issue of the PT-INR in guiding procoagulant therapy is the recognition that this test bears little relationship to actual bleeding risk. Alternative means of assessing bleeding risk in these patients is an essential clinical goal which warrants high priority in clinical trials. Until the time when sufficient data is available, the clinician is left with clinical judgment (such as assessment of prior bleeding events) and the recognition that unnecessary prophylactic plasma may produce unforeseen problems (such as TRALI or exacerbation of portal hypertension) in these frail individuals.
The PT to monitor the treatment with VKA in patient with liver disease
The restored balance of coagulation which results from the parallel reduction of pro- and anti-coagulant proteins may explain the occurrence of hypercoagulability and thrombotic complications in patients with liver disease (reviewed33). Therefore, the accepted notion that the coagulopathy defined solely on the basis of the prolonged PT may indicate auto-anticoagulation should be revised. Patients with stable liver disease may be regarded as ‘normal’ subjects who may occasionally be exposed to circumstantial, or genetic42 risk factors leading to thrombosis. When this happens, liver disease patients may benefit from prophylaxis with VKA. This area has not yet been thoroughly explored and one may wonder which should be the therapeutic interval of anticoagulation to be used in these patients and more importantly which should be the test for dose-adjustment of VKA. As noted, the PT is prolonged at baseline in these patients and therefore may or may not reflect the true level of anticoagulation achieved by VKA. Reasonable solutions to this problem may come from newer antithrombotic drugs which can be given at a fixed dosage without laboratory control.43 One of such drugs as the oral thrombin direct inhibitor ximelagatran proved effective in comparison with VKA in many clinical settings,43 but, ironically, it was withdrawn from the market because of liver toxicity. Other drugs acting against thrombin or factor Xa are under development43 and may be suitable alternatives to VKA in the future.
Use of the PT and the INR scale to prioritize patients for liver transplantation
The MELD score is a mathematical function which includes bilirubin, creatinine and the PT with results expressed as INR.44 The MELD score was originally developed as a means of predicting survival after creation of transjugular intrahepatic portosystemic shunt in cirrhotics.8 However, in response to the problems which developed in donor liver allocation, the MELD score was adapted to liver transplantation prioritization. Prior to the institution of MELD-based organ allocation, the distribution of livers was based, in part, on subjective factors. As a result, there was a perception that widespread subjective ‘upgrading’ had corrupted the integrity of the national liver allocation system.45 Consequently, the MELD score was developed ostensibly to prioritize patients solely on ‘objective’ laboratory parameters. In fact, according to the United Network for Organ Sharing documents, the MELD score is ‘simple, objective and verifiable and yields consistent results whenever the score is calculated’.46 However, there are numerous studies demonstrating that the PT-INR when used in patients with liver disease has substantial inter-laboratory variation47–51 (reviewed in Table 2) and therefore may not be as ‘objective’ a parameter as commonly thought. This variation, which applies to patients with liver disease, but not to patients receiving VKA translates in a significant difference in MELD score (Table 2) and result in substantial changes in prioritization for liver transplantation. The wide variations in PT-INR for liver patients compared to patients receiving VKA are not surprising. They are related to the principle of thromboplastin calibration and to the mechanism of PT prolongation caused by VKA as opposed to liver disease (see previous sections). As noted, the INR scale was never designed for its current use within the MELD score and its adoption in this context is inappropriate and misleading because it does not secure harmonization of results obtained in different laboratories. Although the PT test per se may be regarded as a reliable prognostic index, the expression of its results as INR in patients with liver disease may undermine the fundamental purpose of the current allocation system which is to achieve national parity for transplantation and to prioritize patients solely on objective measures.
Table 2. Review of studies which showed inter-laboratory differences in the international normalized ratio (INR) as measured with different thromboplastins for patients with liver disease vs. patients on vitamin K antagonists (VKA). The resultant inter-laboratory difference in the MELD score is shown in brackets
Mean INR (MELD) % differences (from the highest to the lowest)
Recent studies have cast doubt on the validity of many of the assumptions that form the basis for the application of the PT to assess bleeding risk and prognosis in patients with liver disease as well as to guide the management of the associated coagulation disturbances. It is our hope that this paper will raise awareness of these problems and generate the research required to solve these clinical challenges. Here some directions on how to proceed.
Bleeding risk assessment
The PT, which measures only the activity of a few procoagulant factors, fails to capture changes in other components necessary for hemostasis as well as changes in anticoagulant activity that occur in liver disease. It is therefore not surprising that changes in the PT do not reliably predict the risk of bleeding. These data underscore the need for new tests more suitable than the PT to evaluate bleeding risk in this population. Thrombin generation monitored by means of test systems which mimic more closely in vivo conditions16 might be more appropriate than traditional tests. Furthermore, thromboelastography may give relevant information for the assessment of bleeding risk in liver disease patients.52 Although providing a much more comprehensive examination of the pro- and anti-coagulant pathways, these tests need to be prospectively evaluated in clinical trials to assess their relationship to bleeding.
Monitoring anticoagulant therapy
The problem of managing cirrhotic patients who develop or possess high risk of thrombosis remains a challenge which requires much further work. Conventional and newer emerging anticoagulants which may require less monitoring in non-cirrhotic patients43 will require clinical investigation in the liver diseased patient to determine their net effect. As with bleeding risk assessment, we feel that such tests that take into account the action of both pro- and anti-coagulants as the thrombin generation test performed in the presence of thrombomodulin,16 or thromboelastography52 offer a much more valuable means of assessing this problem.
Prioritizing patients for liver transplantation
The potential for misusing the inter-laboratory variability of the INR to ‘game’ the MELD system for organ allocation should drive the development of methods able to minimize such differences or alternate algorithms for organ allocation that are less susceptible to variability. The theoretical utility of the PT might be achieved only if it could be standardized for liver disease in a manner analogous to its standardization for the management of oral anticoagulation. Recently, the feasibility of an alternative system of thromboplastin calibration has been evaluated.53 This system calls for the determination of ISI values for commercial thromboplastins by using plasmas from patients with cirrhosis instead of plasmas from patients on VKA.53 This alternative calibration proved reliable for the normalization of PT across thromboplastins, but its implementation requires an international effort and partnership with industry. Alternatively, adoption of other expressions of results for the PT such as the percentage activity, which proved effective in harmonizing PT results in patients with liver disease49 or alternative prognostic indices such as some measure of portal pressure less affected by inter-laboratory variation should be explored.
Declaration of personal and funding interests: None.