Updated MELD Model Improves Prediction of Organ Wait-List Mortality

Jim Kling

May 11, 2010

May 11, 2010 (San Diego, California) — The Model for End-Stage Liver Disease (MELD) score for liver transplant recipients successfully predicts mortality in patients who are wait-listed, but refinements have been proposed ever since it was adopted. A new study, presented here at the American Transplant Congress 2010, shows that the addition of serum sodium levels and the inclusion of lower and upper bounds (LB and UB) of some variables can improve the predictive capability of MELD.

The Scientific Registry of Transplant Recipients (SRTR) proposed updated coefficients on the basis of wait-list data from 2005 and 2006 (SRTR MELD). The researchers set out to determine if adding sodium and LB and UB to variables would improve SRTR MELD.

The researchers used wait-list data from the Organ Procurement and Transplantation Network from 2005 to 2007, and divided them into a model derivation set (2005 and 2006) and a model validation set (2007). The analysis encompassed all adult primary liver transplant candidates with end-stage liver disease, and involved 20,906 wait-list registrants — 13,694 in the model development set and 6,942 in the validation set. The researchers used Cox proportional hazards regression analysis to examine the UB and LB of sodium and the existing SRTR MELD variables.

Analyses showed that the addition of UB and LB yielded a significant improvement in the model for international normalized ratio (INR), creatinine, and sodium, but bilirubin did not benefit from a UB.

The researchers identified optimum values for each variable: bilirubin (LB, 1.0 mg/dL; UB, none); creatinine (LB, 0.8 mg/dL; UB, 3.0 mg/dL); INR (LB, 1.0; UB, 3.0); and serum sodium (LB, 125 mEq/L; UB, 140 mEq/L).

"The data show a natural upper and lower cap, which should be incorporated [into the MELD model]. An upper cap in INR is a new concept, and it should decrease concerns about lab-based variability. . . . [Our bounded] MELD demonstrates the best model fit and when we add sodium to this formula it shows a modest statistically significant benefit in predicting 90-day mortality, compared with the existing MELD and SRTR MELD models," Michael Leise, MD, gastroenterologist at the Mayo Clinic in Rochester, Minnesota, said during the presentation.

"It's a good study. Everybody is trying to update MELD, and I think their modeling is good," Devin Eckhoff, MD, professor of surgery at the University of Alabama, Birmingham, who moderated the session, told Medscape Transplantation.

The study did not receive commercial support. Dr. Leise and Dr. Eckhoff have disclosed no relevant financial relationships.

American Transplant Congress (ATC) 2010: Abstract 68. Presented May 2, 2010.