Get started
Work

Reply to: "The decreasing predictive power of MELD in an era of changing etiology of liver disease"

Year: 2019
Type: letter
Abstract: To the Editor: We read with interest the recent article by Godfrey et al1Godfrey EL Malik TH Lai JC et al.The decreasing predictive power of MELD in an era of changing etiology of liver disease.Am J Transplant. 2019; 19: 3299-3307Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar regarding the predictive power of the Model for End-Stage Liver Disease (MELD), the score used for allocation of donor livers, over time. The authors report that the discriminative ability of MELD for the prediction of 90-day waitlist mortality, as measured by the concordance statistic (c-statistic), had declined from 0.80 in 2002 to 0.70 in 2015 among new waitlist registrants. These values are inconsistent with previously published estimates.2Leise MD Kim WR Kremers WK Larson JJ Benson JT Therneau TM. A revised model for end-stage liver disease optimizes prediction of mortality among patients awaiting liver transplantation.Gastroenterology. 2011; 140: 1952-1960Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar On review of the statistical analysis in this article, we learn that the c-statistic was calculated using a binary method for 90-day mortality without inputs for contributed follow-up time. In such an analysis, the question asked is whether a given patient is alive or dead as of day 90. The proper method incorporating the length of survival information would consider censored time, which informs whether a patient with a given MELD is more likely to die before another patient with a lower MELD.3Uno H Cai T Pencina MJ D’Agostino RB Wei LJ. On the C-statistics for evaluating overall adequacy of risk prediction procedures with censored survival data.Stat Med. 2011; 30: 1105-1117Crossref PubMed Scopus (756) Google Scholar Consideration of censored time is particularly crucial in an urgency-based allocation system. In Figure 1A, we apply the analysis to the same dataset using a modified (Harrell’s) c-statistic, which accounts for censoring and corresponds to methods previously used in the development and validation of MELD and MELD-Na.2Leise MD Kim WR Kremers WK Larson JJ Benson JT Therneau TM. A revised model for end-stage liver disease optimizes prediction of mortality among patients awaiting liver transplantation.Gastroenterology. 2011; 140: 1952-1960Abstract Full Text Full Text PDF PubMed Scopus (107) Google Scholar,4Kamath PS Wiesner RH Malinchoc M et al.A model to predict survival in patients with end-stage liver disease.Hepatology. 2001; 33: 464-470Crossref PubMed Scopus (3746) Google Scholar,5Kim WR Biggins SW Kremers WK et al.Hyponatremia and mortality among patients on the liver-transplant waiting list.N Engl J Med. 2008; 359: 1018-1026Crossref PubMed Scopus (935) Google Scholar Using the survival package in R, the discriminative power of MELD is 0.84 and has varied minimally since its implementation in 2002. In Figure 1B, results by Godfrey et al are closely reproduced using the pROC package for the logistic c-statistic, ignoring time-to-event data. Comparison of the 2 panels suggests that the decreasing c-statistic in Figure 1B is a result of an increasing number of deaths within 90 days, which the binary method considers to be equivalent whether the death occurred on day 1 or day 89. Undoubtedly, the demographics of liver disease have changed since 2002, and there may be variables not included in MELD that may improve prediction of waitlist mortality. Nonetheless, the predictive performance of MELD endures, exceeding that of many predictive scores used in modern clinical medicine. The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.
DOI: 10.1111/ajt.15733
Language: English
Subfield: Hepatology
Field: Medicine
Open Access status: hybrid
Estimated APC paid: $4,300
OpenAlex ID: W2992223635
Cites: 5
Cited by: 7
Related to: 10