Statistical Analysis

Descriptive statistics were performed with chi-squared testing for categorical variables. Median and interquartile range (IQR, 25^th-75^th percentile) were reported for continuous variables because of skewed distributions. Kruskal-Wallis testing for between group differences was utilized for continuous variables. Mantel-Haenszel testing was used for unidirectional trend in analyses of NSE denial by listing year.

Risk of wait list death, defined as a death or waitlist removal for being too sick to transplant, was evaluated using Fine and Gray competing risks regression.(8) Observation time was measured from the date of listing for transplant to waitlist death (event), liver transplant (competing risk), or last date on the waiting list for patients still waiting or removed for other reasons (censored). Risk of wait list death (subhazard ratios, SHR) was estimated by modeling the cumulative incidence function with competing risks regression for demographic and clinical characteristics. Factors with p<0.1 in the univariate analysis were evaluated in the multivariable model. Factors were eliminated by backward stepwise selection; NSER, as the primary predictor of interest, and all other variables that retained p<0.05, were reported in the final model.

Risk of post-transplant outcomes, patient death and graft loss (defined as re-transplant or death), within 3 years of transplant were evaluated using Cox proportional hazards regression. (9) Patient follow-up time was measured from the date of transplant to the first event, retransplant (when graft loss evaluated) or death, or last follow-up within 3 years of transplant (censored). Univariate hazard ratios (HR) estimated risk of patient death or graft loss for demographic and clinical characteristics. Factors with p<0.1 in the univariate analysis were evaluated in the multivariable model. Factors were eliminated by backward stepwise selection; NSER, as the primary predictor of interest, and all other variables that retained p<0.05 were reported in the final model.

Data analysis was completed using SAS version 9.4 (SAS Institute, Cary, NC) and Stata/IC 14 (StataCorp, College Station, TX).