How much is too much?^†

Ascites is the most common complication of portal hypertension associated with cirrhosis. The development of ascites is a harbinger of a poor prognosis and impaired quality of life. The mortality rate is approximately 50% 2 years after the development of ascites.1 Independent factors associated with ascites-related mortality include hyponatremia, increased serum creatinine levels, low arterial pressure, and low urine sodium levels.1, 2 Among these variables, only the serum creatinine level is part of the formulation of the Model for End-Stage Liver Disease (MELD), which predicts 90-day mortality for patients awaiting liver transplantation. Therefore, the development of a risk prediction model to further improve risk profiling, especially for patients on the transplant waiting list, remains a top priority. Unfortunately, ascites is a subjective clinical marker that is difficult (if not impossible) to incorporate into defined, objective survival models. Heuman et al.3 previously demonstrated that hyponatremia and persistent ascites are MELD-independent predictors of early mortality and are especially important in patients with MELD scores lower than 21.

Abbreviations:

DRI, donor risk index; MELD, Model for End-Stage Liver Disease; MELD-Na, Model for End-Stage Liver Disease plus serum sodium.

In the field of liver transplantation, the greatest challenge continues to be the shortage of donor organs. The landscape has morphed and now includes the risk of using extended criteria donors because this might be the only opportunity or chance for patients. Therefore, developing an accurate, objective survival model that is fair and totally inclusive in association with donor allocation may be as difficult as total health care reform! In this issue of Liver Transplantation, Somsouk et al.4 address the daunting question of ascites and mortality risk while patients wait for liver transplantation.

Using the Organ Procurement and Transplantation Network database, Somsouk et al.4 examined all new registrations for liver transplantation from 2005 to 2007 with follow-up to 90 days. Patients who were removed from the waiting list and died later were also considered to have died while they were waiting. Ascites was entered into the database by transplant coordinators at their centers at the time of registration and was classified as none, small, or moderate; 57% were noted to have small ascites, and 25% were noted to have moderate ascites. The study group consisted of 18,124 patients, and 1498 (8.3%) died. The mortality rate was greater in patients with moderate ascites versus those with no or small ascites (15.4% versus 4.1% and 15.4% versus 6.6%, respectively). With adjustments for the MELD score, the risk of death was doubled for patients with moderate ascites. Furthermore, in comparison with MELD and Model for End-Stage Liver Disease plus serum sodium (MELD-Na) scores, moderate ascites offered additional risk discrimination for predicting 90-day mortality. Somsouk et al. found that the mortality rate was higher in patients with moderate ascites, and the effect was more prominent with MELD scores lower than 21 (equal to 4.7 MELD units) and with MELD-Na scores lower than 21 (equal to 3.5 MELD-Na units). Lastly, the risk of death in high-demand US allocation regions for patients with MELD scores lower than 21 and moderate ascites was 8% higher than the risk in lower demand regions.

There are several key points to this article.4 These data confirm previous studies reporting that moderate ascites is an independent risk factor associated with death, especially in patients with low MELD scores. Moreover, the authors quantify what this risk is with respect to risk-adjusted MELD scores. Lastly, patients with moderate ascites in high-demand regions have higher wait-list mortality.

There are limitations to this study.4 Moderate ascites is defined as a subjective variable in this study and remains ill defined by the International Ascites Club (moderate ascites evident by moderate symmetrical distension of the abdomen).5 The data were entered by a liver transplant coordinator, most likely after a review of a clinical assessment note or an ultrasound report, and this leaves interpretation bias on the table. The data were registration data at the time of listing for transplantation; therefore, it is unknown how many patients, if any, had progressive or difficult-to-control ascites during their wait-list follow-up, and information about diuretic use and dosage was not available. However, because 25% of the patients were considered to have moderate ascites, it is very unlikely that most of these patients had refractory ascites (the median survival time was approximately 6 months). The causes of death for the patients were not available. I wonder whether most of the deaths were related to complications of ascites, such as spontaneous bacterial peritonitis and the development of hepatorenal syndrome. Lastly, even if patients are told that their risk of death is higher than the MELD-calculated risk because of moderate ascites, the use of extended criteria donors in this particular group in high-demand regions may or may not increase their chances for transplantation. A review of United Network for Organ Sharing regional variation by the donor risk index (DRI) for 2006-2008 showed uniformity in all regions except region 9 (1.37 in region 1, 1.54 in region 2, 1.50 in region 3, 1.37 in region 4, 1.37 in region 5, 1.37 in region 6, 1.48 in region 7, 1.43 in region 8, 1.86 in region 9, 1.45 in region 10, and 1.42 in region 11; S. Feng, personal communication, 2010). Therefore, patients in high-demand regions do not seem to be gaining increased access to high-DRI livers in comparison with patients in low-risk regions.

Why do patients with MELD scores lower than 21 and moderate ascites have a higher risk of death? I believe that this is most likely related to the high portal pressure in these patients, which is not reflected in the MELD score. Investigators from Spain have noted that the hepatic venous pressure gradient is an independent variable predicting death: each 1 mm Hg increase in the hepatic venous pressure gradient predicts a 3% increase in the risk of death!6 A major breakthrough would be the development of a noninvasive, objective variable that reliably reflects the portal pressure and could be incorporated into survival/allocation models. As noted by the authors,4 further studies are needed to define an objective definition of moderate ascites. In addition, the authors do not propose the replacement of the current MELD allocation system with a subjective variable of ascites; the elimination of subjective variables was one of the intents of the MELD system in the first place.

Multiple reports have noted an inverse correlation between the DRI and the MELD score.7, 8 Therefore, one premise of this study seems to be in line with current clinical practice: high-DRI donors (expanded criteria donors) may be appropriate for patients with moderate ascites and low MELD scores. In contrast, others believe that high-DRI livers yield better survival benefits in candidates with MELD scores greater than 20.9

How do we use this information? Yes, I believe that patients with moderate or difficult-to-control ascites (especially those with hyponatremia) have a higher risk of death not reflected by the MELD score. With the current US-based MELD allocation system, the only choice is to apply for exception points to increase the MELD score to reflect a patient's true mortality risk. Currently, 30% of patients on the waiting list receive MELD exception points; however, exception points are rarely granted for ascites. This study estimates an appropriate MELD increase and might increase the probability of acceptance for exception points. I do not believe that the transplant community wants to go back to including subjective criteria as part of the allocation formula. Informing patients of their true risk of death on the waiting list and the inherent risk of potential donors (eg, high-DRI donors) should be standard practice for any transplant program.