Reply

Abstract

As Dr. Russo highlights, we found serum sodium, at the date of listing for liver transplantation and at any time while on the wait list, to be strongly associated with wait-list mortality, independent of the MELD score.1 For a serum sodium <126 mEq/L at any point while on the wait list, the lowest excess risk consistent with our results was 2.9 times that of those with higher serum sodium concentrations (hazard ratio of 6.26, 95% CI 2.92-13.4). Serum sodium as a continuous variable also had a significant association with wait-list mortality with 7% increase risk of wait-list mortality for each unit decrease in listing serum sodium. Heuman et al. found that the association with serum sodium and wait-list mortality was limited to patients with MELD <21.2 When we examined the effect of different MELD scores (<16, 16-30, and >30) on strength of association between the serum sodium and wait-list mortality, serum sodium as a continuous variable and as a dichotomous variable (< or ≤126 and > or ≥131), at listing and at any time point, the serum sodium variables were statistically significant across the spectrum of MELD scores. There was no specific cutoff MELD score identifiable that resulted in loss of statistical significance of serum sodium as a predictor of mortality. Logistic regression models provided another way to evaluate the association between hyponatremia and wait-list mortality at 3 and 6 months. In comparison to MELD alone, the addition of serum sodium <126 mEq/L as a dichotomous variable plus MELD led to an increase of the accuracy (c-statistic) of predicting 3- and 6-month wait-list mortality by 3.4% (from 0.883 to 0.917) and 5% (from 0.871 to 0.921), respectively. The estimated incremental improvement with the addition of serum sodium to MELD was sizable considering the shift from CTP to MELD for organ allocation was largely based on a 7% change in the c-statistic (from 0.76 to 0.83) reported by Wiesner et al.3 Russo indicates that the P values comparing ROC curves for MELD alone versus MELD plus serum sodium did not achieve statistical significance. This is true, but the c-statistic represents only one way to assess for significance. The uniform increase in area under the ROC curves (c-statistic) with the additional of serum sodium to the model and the significant increase in odds ratios in the logistic regression models are indicative of a significant independent impact of serum sodium predicting wait-list mortality. Moreover, an equally important measure of the statistical significance of the adding serum sodium to MELD is reflected by the 95% confidence intervals for the bivariate logistic regression models that generate the ROC curves. These results showed that the increase in odds of 3- and 6-month wait-list mortality, controlling for MELD score, was at least 4% and 5%, respectively, for each unit decrease in listing serum sodium concentration. Russo comments that the model by Heumann et al. is more “encouraging,” possibly because the findings of that study are in keeping with his clinical intuition. However, our conclusions that serum sodium adds to MELD in predicting wait-list mortality is strongly supported by the survival analyses and logistic regression models. Clearly, in the future, we look toward confirmation and extension of our findings in larger, multicenter databases. With UNOS currently collecting serum sodium values, we anticipate such studies will be forthcoming in the very near future. Scott Biggins M.D.*, Norah Terrault M.D., M.P.H.*, * University of California San Francisco, San Francisco, CA