Descriptive statistics (means, standard deviations and proportions) for various patient characteristics were calculated separately for patients who eventually received an LDLT, a DDLT or no transplant. For LDLT and DDLT recipients, explant tumor characteristics were also summarized. Statistical differences between the LDLT and DDLT recipients’ characteristics and explant tumor characteristics were evaluated using chi-square and t-tests. A cumulative incidence function to display the probability of receiving LDLT or DDLT, dying on the waitlist, and remaining alive without transplant over time since donor evaluation (17) was calculated using the SAS macro ‘comprisk’ developed at Mayo Clinic (18). Freedom from HCC recurrence, recurrence-free survival and patient survival were estimated by Kaplan-Meier methods and unadjusted comparisons made using the log-rank test. In the case of complete separation (no HCC recurrence in DDLT), the effect of LDLT was tested by Cox regression likelihood ratio test. Potential predictors of time from transplant to death, HCC recurrence or the combined endpoint were tested by fitting multivariable Cox regression models. Covariate effects were reported as hazard ratios (HR) with 95% confidence intervals (CI). A wide range of covariates were evaluated. Recipient variables included age, gender, race, etiology of underlying liver disease, calculated MELD score at transplant, HCC staging at enrollment, pretransplant HCC tumor ablation and/or chemotherapy, and AFP at listing, enrollment and transplant; donor and transplant-related variables included donor age, LDLT case number (10), transplant era, time from listing to transplant, time from donor evaluation to transplant, explant characteristics (stage, grade and vascular invasion), transplant center, year of transplant and cold ischemia time. We also investigated whether there was a survival advantage for HCC candidates who pursued LDLT, comparing the relative mortality risk for LDLT recipients to that of waitlisted candidates including those who subsequently underwent DDLT. A modified Cox regression model of time from first donor evaluation to death was fitted, using the method of sequential stratification (17). Briefly, for all LDLTs performed at a given number of days since first donor evaluation, a separate comparison group (stratum) was created that included all patients alive and without transplant prior to the time of the index LDLT(s). The survival of the index LDLT patient(s) was compared to that of the other patients in that stratum, and the results across all LDLT strata were pooled in a stratified Cox regression. The variance was adjusted to account for patients who were in the comparison groups of multiple strata. Patients without LDLT at entry into a stratum were censored if they later received an LDLT. All analyses were performed using SAS version 9.1 (SAS/STAT 9.1 User's Guide, SAS Publishing, Cary, NC: SAS Institute Inc., 2004). P values <0.05 were considered to be statistically significant.