We retrospectively analyzed 1856 consecutive patients (69.3% male, mean 66.8 ± 12.2 years) undergoing cardiac surgery between 2008 and 2013 at our institution.
The mean values of MELD/MELD-XI scores obtained from the total cohort were 9.7 ± 4.6/11.9 ± 4.8. Patients with high MELD/MELD-XI scores (> 12) were older, more anemic, and had lower left ventricular ejection fraction than those with low scores (all p < 0.0001). High scores were associated with longer hospitalization (36.8 ± 33.0 vs. 23.6 ± 21.5 days for MELD, 36.7 ± 34.0 vs. 23.5 ± 21.0 days for MELD-XI, both p < 0.0001) and higher in-hospital all-cause mortality (6.6% vs. 1.0% for MELD, 7.7% vs. 0.7% for MELD-XI). Atrial fibrillation occurred more frequently in the high MELD group, but this difference was not found for MELD-XI. MELD/MELD-XI could predict mortality with a sensitivity of 64.3%/60.0% and specificity of 79.1%/85.3%. Comparison of AUC values among MELD scores, individual MELD components, and Child–Turcotte–Pugh (CTP) classification showed that the predictive strength of MELD scores for mortality was stronger than individual parameters or CTP classification (AUC: 0.7702 for MELD, 0.7655 for MELD-XI, 0.5799 for CTP classification with pairwise p < 0.0001 and p = 0.0002 vs. MELD and MELD-XI, respectively).
Assessment of liver dysfunction using the MELD scores can be useful for predicting postoperative morbidity and mortality, which may allow additional risk stratification in patients undergoing cardiac surgery.