Higher Mortality in Patients With Sudden MELD Increase: Disadvantaged By the Current Allocation Policy.

Abstract

Priority of allocation on the deceased donor liver waitlist is based on current MELD; ties are broken by waiting time accrued at a current or higher MELD value. However, registrants whose MELD has recently increased may be at higher, not lower, mortality risk. We modeled 7-day mortality risk in 67141 adult liver waitlist registrants from 2002-2013 using logistic regression. Using a likelihood maximization procedure, we defined a MELD “spike” as a MELD increase of at least 30% over the previous 7 days. 38.9% of patients had at least one spike while on the liver waitlist. Registrants with a spike in the previous 7 days had higher risk of mortality in the following 7 days than registrants at the same current MELD with no spike (Figure 1A). The additional risk of death conferred by a spike was equivalent to 2-9 additional MELD points depending on current MELD: specifically, a registrant with a spike and a MELD of 10, 20, or 30 had mortality risk equivalent to those with a MELD of 18, 26, or 34 and no spike. Risk increase associated with a spike was highest for patients with low MELD scores (Figure 1B): specifically, 3.77-fold (3.59 - 3.95), 2.42-fold (2.35 - 2.49), 1.55-fold (1.51 - 1.60) higher odds of 7-day mortality at MELD of 15, 25, and 35.Figure: No Caption available.Figure: No Caption available.In a calibration plot (Figure 2) of predictive validity among patients with a recent MELD spike, predicted mortality risk from a model (model 1) with no spike was consistently less than observed risk, while predicted mortality risk from a model (model 2) incorporating spike matched observed risk more closely.Figure: No Caption available.Registrants with a recent sudden MELD increase have higher risk of short-term mortality than indicated by their current MELD. As such, allocation policy should be adjusted to reflect this risk.