In a MELD-based economy, how can we fight off inflation?

Since the implementation of the Model for End-Stage Liver Disease (MELD) for deceased donor liver allocation, the MELD score has become the gold standard for clinicians to communicate the severity of liver disease. Most transplant physicians have developed a “feel” for what a patient with a MELD score of 10 or 36 will “look like.” Yet, since some complications of chronic liver disease do not manifest in derangements of the MELD score, alternative prioritization via MELD exceptions has been permitted. A prime example is hepatocellular carcinoma (HCC). Candidates with HCC are assigned transplantation priority based on their predicted risk of “drop out,” that is, progression to a tumor stage with unacceptable rates of posttransplantation recurrence.1 Iterative evidence-based adjustments have established an exchange rate between the risk of HCC “drop out” and the risk of death from liver disease. In this manner, MELD points have essentially become the currency for transplantation priority. MELD, Model for End-Stage Liver Disease; HCC, hepatocellular carcinoma; MELD-XI, MELD excluding international normalized ratio; INR, international normalized ratio. As with any currency, there is the risk of inflation. With regard to MELD allocation, inflation effectively raises the threshold at which liver transplantation occurs. MELD exceptions have been identified as potential drivers of MELD inflation.2 HCC, the leading MELD exception, has already undergone 2 devaluations. Unlike HCC, however, many other conditions associated with chronic liver disease have not had a predetermined MELD value set by national policy. Rather, these conditions have been variably valued, reflecting local market forces by regional review boards. With HCC as the benchmark, the International Liver Transplantation Society recently convened the single topic consensus “MESSAGE” meeting in March of 2006 to discuss national guidelines to specify the conditions that would be eligible for MELD exception and the algorithms by which their MELD exchange rate would be set. In this issue of Liver Transplantation, Heuman et al.2 report on a novel mortality risk model for cirrhotics, MELD excluding international normalized ratio (MELD-XI), which may serve to stave off another potential source of inflationary (and even deflationary) pressure on MELD, artificial alterations in the international normalized ratio (INR). Using a dataset comprised of 554 cirrhotic veterans referred for liver transplant consideration, the authors derived MELD-XI, which uses 2 of the 3 MELD parameters, serum total bilirubin and creatinine. MELD-XI was validated using a “holdout” dataset (a subset of the original dataset) and an independent dataset of pre-MELD era candidates from the United Network for Organ Sharing. Importantly, they normalized MELD-XI to the standard MELD scale, thereby retaining the mortality risk connotations with which clinicians have become intimately and intuitively familiar. The authors report that omitting INR from MELD slightly diminishes the predictive accuracy of short-term mortality. Compared to MELD, which had an accuracy of 85.0% to predict 90-day mortality of adult cirrhotics awaiting first liver transplantation, MELD-XI had an accuracy of 83.5%. MELD-XI has several potential important applications. For therapeutically anticoagulated liver transplant candidates, MELD-XI may reflect waitlist mortality more accurately than MELD. Intuitively, the INR of candidates taking warfarin does not reflect hepatic synthetic function. Although indications for warfarin use in cirrhotics are few (i.e., Budd-Chiari syndrome, thromboembolism prevention, hypercoagulable states), the impact of MELD inflation for even a small group can be substantial. Relative to other waitlisted candidates, inflated mortality assessment of even a single candidate represents an unfair advantage in the competition for deceased donor liver grafts. Moreover, for the individual, this inflated mortality assessment distorts the estimated potential survival benefit of liver transplantation. Take a cirrhotic patient with a MELD of 9 (INR = 1.0; total bilirubin = 1.5 mg/dL; creatinine = 1.1 mg/dL). Therapeutic anticoagulation with warfarin to an INR of 3.0 would inflate his MELD to 21 (INR = 3.0; total bilirubin = 1.5 mg/dL; creatinine = 1.1 mg/dL). While liver transplantation would offer a significant survival benefit for a cirrhotic patient with a MELD of 21, it would not for a cirrhotic patient with a MELD of 9.3 In fact, for a candidate with a MELD of 9, the 1-yr risk of death with transplant is 3.6 times higher than that of remaining on the waitlist.3 Here, application of MELD-XI would assist the therapeutically anticoagulation cirrhotic patient and his physicians to better evaluate the decision to proceed with or decline an opportunity for transplant. If MELD-XI, compared to MELD, provides a more appropriate assessment of waitlist mortality risk in candidates who require anticoagulation, what about candidates with chronic liver disease who require reversal of coagulopathy? Unlike therapeutic anticoagulation, correction of coagulopathy is typically short-lived, in the context of an invasive procedure or acute event. Since a nonfulminant candidate's highest MELD score in the preceding 7 days is used by United Network for Organ Sharing for allocation, the MELD score calculated before correcting coagulopathy could generally (but not always) be expected to reflect the estimated mortality. Candidates with fulminant hepatic failure, listed as status 1, are prioritized independently of their MELD score and therefore allocation is not affected by correction of their coagulopathy. If MELD-XI is nearly as accurate as MELD in predicting waitlist mortality, should we just replace MELD with MELD-XI? Despite recent reports citing the interlaboratory variation of INR as a source of systematic bias in calculating MELD score,4 numerous studies including the current report by Heuman et al.2 have demonstrated that increasing INR is a strong, independent predictor of mortality with liver disease.2, 5-7 Considering the exhaustive validation of MELD in a broad spectrum of liver diseases5-8 and its general success as an allocation tool,9-11 it would seem premature and unwise to switch to MELD-XI. The triad of parameters, INR, total bilirubin, and creatinine, diversifies the predictive capacity of MELD across the major manifestations of decompensated liver disease, coagulopathy, cholestasis, and renal dysfunction. INR has been shown to be the dominant predictor of mortality for candidates with fulminant hepatic failure12, 13 and for candidates with recurrent hepatitis C virus cirrhosis after liver transplantation.14 MELD-XI was derived excluding these subpopulations and may not stand up to broader application. For optimal mortality prediction, diversification in the MELD parameters, like in a good stock portfolio, is a wise strategy. The MELD-based allocation system exemplifies the liver transplant community's commitment to distribute deceased donor livers to candidates according to objective assessment of medical urgency. Without doubt, additional adjustments will follow. Vigorous scrutiny of MELD has already yielded several beneficial evidence-based adjustments in MELD valuation that have focused predominantly on conditions associated with risk that are not reflected or captured by the calculated MELD.11 The work of Heuman et al.2 highlights the fact that, for specific candidate subsets, however, the laboratory value for 1 or more of individual components of MELD may not accurately reflect liver disease severity for specific candidate subsets. This concept has previously been acknowledged relative to bilirubin, for which cholestatic and noncholestatic diseases were differentially valued in the Child-Turcotte-Pugh classification15 and in the original derivation of MELD.16 We believe that there is a pressing need to explore the third component of MELD, creatinine, particularly in the contrasting settings of acute renal dysfunction secondary to hepatorenal syndrome or acute tubular necrosis vs. chronic renal dysfunction secondary to intrinsic renal disease. Currently, a candidate with normal liver function (total bilirubin = 1.0 mg/dL, INR = 1.0) on dialysis for intrinsic renal disease (creatinine set at 4.0 mg/dL) has a calculated MELD score of 20. However, this candidate's short-term mortality may well approximate that of a dialysis patient awaiting renal transplantation (9.02 per 100 patient-yr).17 In contrast, the decompensated cirrhotic with a calculated MELD score of 20 faces a much higher short-term mortality risk (27.2 per 100 patient-yr).3 Investigations to improve mortality estimates in cirrhotics with chronic vs. acute renal insufficiency or failure should be strongly encouraged. As advocates for our individual patients, there is a natural tendency for transplant physicians to try to maximize MELD scores and thereby improve the access of our patients to deceased donor livers. Yet as custodians of this scarce resource, we must refrain from contributing to MELD inflation—raising the MELD score required to achieve transplantation. MELD is robust predictor of short-term mortality across a broad spectrum of liver diseases and serves as an excellent measure of liver transplant urgency for the vast majority of waitlist candidates. Although MELD sets a high bar, further refinements in the MELD-based allocation system are clearly possible and should be aggressively pursued in an evidence-based manner.