The prothrombin time‐derived international normalized ratio: great for Warfarin, fair for prognosis and bad for liver‐bleeding risk

Like a politician who cannot or will not let go of their post, the standard prothrombin time (PT)-derived international normalized ratio (INR) test persists, as yet unchanged and as a daily-used tool in critical decision making among liver disease patients – this, in spite of absorbing repeated broadsides that have removed its masts and revealed the numerous limitations inherent to a test created to evaluate vitamin K antagonist (VKA) therapy, but applied then to patients with an entirely different type of coagulation problem. The persistence of the test as a component of the widely used model for end-stage liver disease (MELD) prognostic score likely relates to both the improvement in organ allocation that came with implementation of the MELD system in the USA and to the not insurmountable challenges of implementing changes that can resolve interlaboratory variation (1–3). However, its persistence in assessing bleeding risk relates more to entrenched non-evidence-based (and sometimes ludicrous) dogma and the misperceived lack of alternatives to guide prophylactic therapy (4). In this issue of Liver International, Lisman et al. (5) further address the problem of MELD score reproducibility in a series of 15 patients whose MELD scores were determined from single blood specimens sent to seven different clinical laboratories in Europe. The current paper extends work from Trotter and colleagues by further demonstrating that the problem of INR reproducibility is worldwide and persists in spite of using alternative techniques for measuring the PT – the Quick vs. the Owren method (6, 7). In contrast to the liver disease patients, the same laboratories performed well in the VKA-treated control group. In the liver disease group, however, the resulting differences in MELD scores between laboratories for the same individual were striking and largely owing to the INR with a mean difference of 4.8 MELD points between the highest and the lowest labs, and a lesser contribution of the creatinine measurement, with a mean difference of 1.2 MELD points between the highest and the lowest labs. Moreover, differences between the highest- and lowest-scoring laboratory increased with higher MELD scores. Obviously, variation in the MELD defeats the purpose of the system, which is intended to allocate scarce organs more fairly to those in greatest need. On the other hand, there is no doubt that institution of the MELD scoring system in the USA was an enormous step forward in organ allocation, which had depended previously on the sometimes subjective parameters (ascites and encephalopathy) of the Child–Pugh–Turcott score and the infamous ‘time on the waiting list’ parameter that invited abuse of the system (8). The MELD prognostic score, originally derived for predicting post-transvenous intrahepatic portosystemic shunt survival, was formulated to maximize reproducibility and, indeed, it has withstood a great deal of scrutiny over the years in its application to organ allocation as debated recently by Kamath (9–11). In spite of these achievements, refinements in the system especially with regard to the INR are clearly needed (12). Both the mechanism of INR variation that relates to standardization of the thromboplastin against VKA-treated patients rather than liver disease patients and the means by which this problem can be resolved are now known – work by Tripodi et al. and Bellest et al. (13, 14) have shown clearly that there now remains a question of implementing the INRLiver derived against a reference standard of liver disease patients to measure the international sensitivity index (ISI)Liver of the thromboplastin in place of the ISIVKA and the resulting INRVKA. This change results in reproducibility in liver disease patients comparable to that achieved with VKA therapy. In contrast, the misuse of the INR in bleeding risk assessment remains a very tough nut to crack due in large part to perpetuation of common practices in spite of the lack of supporting data and indeed in the face of evidence to the contrary (15–17). There is no doubt that bleeding is common in liver disease patients; however, the coagulopathy is far more complex than that of a coumadin-treated patient. In a landmark study, Boks et al. (18) demonstrated the inability of any single coagulation parameter to correlate with bleeding in either decompensated cirrhosis or acute liver failure. Indeed, these patients may be hypercoagulable in spite of INR prolongation (19, 20). Current assays fail to detect this possibility or the presence of other relevant processes such as hyperfibrinolysis or dysfibrinogenaemia (21–23). Moreover, thrombin production is relatively normal in cirrhosis regardless of the INR when conditions are controlled for simultaneous impairment of the anticoagulant system (factors C and S and AT III), which is also reduced in liver disease (24). On the other hand, there is evidence that thrombin production is significantly improved by increasing the platelet count to the range of 55 000 or more in cirrhosis (25). In summary, the current paper by Lisman and colleagues, emphasizes the need for broadly refining our collective understanding of coagulation disorders in liver disease and especially how we should go about measuring this. The prognosis for improving the MELD score reproducibility is good as both the issue and the solution have been well framed. On the other, the likelihood that we will have to continue to endure artificial cut-offs for the PT-INR in cirrhosis patients seems high. Hopefully, the presence of up to 31% variation between clinical laboratories in INR, demonstrated in Lisman's paper, will drive another nail into the coffin of the practice of using a spurious and unfounded cut-off value to guide prophylactic therapy. Clearly, clinical investigation is needed to determine how best to measure bleeding risk and whether or not rescue interventions are a superior strategy to prophylactic measures (26).