Meta-Analysis and Meta-Regression of Outcomes for Adult Living Donor Liver Transplantation Versus Deceased Donor Liver Transplantation

Arianna Barbetta; Mayada Aljehani; Michelle Kim; Christine Tien; Aaron Ahearn; Hannah Schilperoort; Linda Sher; Juliet Emamaullee


American Journal of Transplantation. 2021;21(7):2399-2412. 

In This Article

Experimental Methods

Literature Search and Study Selection

This systematic review was performed according to the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and as outlines in a predefined protocol (PROSPERO 2018: CRD42018104794).[23] A health sciences librarian developed the search strategy and searched the following databases on March 28, 2018: PubMed (coverage 1946- present), Embase and Embase Classic (coverage 1947-present), Cochrane Library (coverage 1898–present), Web of Science (coverage 1900-present),, and Google Scholar. No filters were applied for date, study type, language, or any other limit. A combination of subject headings (when available) and keywords were used for the concepts living donor, deceased donor, and liver transplantation. See Table S1 for full search strategies and database details. Duplicated citations were removed in EndNote x9.2 using the Bramer method.[24] Cross-referencing and forward searches of articles fulfilling inclusion criteria were performed using Web of Science.

Study Selection

Screening was independently performed by two authors. Any conflict regarding study inclusion was resolved by the senior author. Studies were included if they were published between January 2005 and December 2017, available in full text, compared LDLT and DDLT cohorts, studied transplant recipients ≥18 years of age, and reported on the primary outcome of overall patient survival at ≥1-year posttransplant. A study was excluded if it was limited to <10 patients, did not include DDLT as a reference group, did not differentiate pediatric recipients from adults, did not report patient demographical information or pretransplant characteristics, or did not describe its methods of statistical analysis. Studies including multi-organ transplants, re-transplants, and those reporting only acute liver failure were also excluded.

At the outset, we anticipated that we would include A2ALL data. The most recent comprehensive analyses of A2ALL recipient outcomes include data from ~1000 LDLT and ~500 DDLT recipients from 11 U.S. centers and Toronto, performed between 1998 and 2010.[25,26] While both studies reported primary outcomes of graft and patient survival, neither included the majority of the secondary outcomes formatted for meta-analysis. Based on the Cochrane Handbook for Systemic Review of Interventions, we ultimately excluded the A2ALL papers and other U.S. single center papers and instead performed a larger, more contemporary SRTR analysis to represent U.S. outcomes, with 2750 LDLT and 58,120 DDLT performed between 2005 and 2017.[27] Two studies from the Toronto collectively describing 193 LDLT and 273 DDLT patients transplanted between 2001 and 2014 were also included, which reported both primary outcome measures and data related to all secondary outcomes.[28,29] Using this approach, we have captured all of the A2ALL centers in this meta-analysis.


A primary, up-to-date analysis of the U.S. SRTR registry data was completed to supplement what is presented in the annual data report, with the intent of including primary and secondary outcomes of interest.[5] For details on the SRTR data and analysis, please refer to Data S1. The data reported here have been supplied by the Hennepin Healthcare Research Institute (HHRI) as the contractor for the SRTR. The interpretation and reporting of these data are the responsibility of the author(s) and in no way should be seen as an official policy of or interpretation by the SRTR or the U.S. Government.

Data Extraction and Outcome Measures

Data extraction from eligible studies was independently conducted by two authors. For all studies, data regarding study design and characteristics (year of publication, first author, country), population characteristics (sample size for each patient cohort, recipients and donor demographics, MELD at transplant), and liver disease diagnosis were recorded when available. The primary study outcome was 1-, 3-, and 5-year patient survival. Secondary outcomes included 1-, 3-, and 5-year graft survival; preoperative variables (MELD score and time on waiting list); and postoperative variables (biliary complications, HAT, infection, rejection, and length of stay).

Assessment of Risk Bias

The assessment for risk of bias was independently carried out by two authors. The NIH Quality Assessment Tool for Case-Control Studies was adopted to evaluate the quality of each included study. Based on the overall score, each study was classified as good (scored 9 or higher), fair (scored between 5 and 8), or poor (lower than 5) (Table S4).

Statistical Analysis

For the meta-analysis, percentage and total numbers were used to report categorical variables and mean with standard deviation (SD) for continuous variables. When included studies reported median and interquartile range, mean and SD were estimated according to established methods.[30] For pooled analyses, all variables reported in ≥5 studies were analyzed. Continuous variables were analyzed by mean difference (MD), whereas categorical variables were analyzed by odds ratio (OR), both with 95% confidence intervals (CI). Random effects model was adopted to balance intrinsic heterogeneity and effect size.[31] Heterogeneity was also assessed with chi-square statistic and I2 statistic with I2 ,≥ 50% representing significant heterogeneity. The hazard ratio (HR) for time-to-event outcomes was estimated indirectly from other summary statistics or from data in published Kaplan-Meier curves.[32] The derived observed minus expected number of cases (O-E) and the variance for the single studies were then used to calculate individual and overall HR with the fixed-effect model to give a pooled HR for survival analyses.[33] Forest plots were created to display results. All data analyses were conducted using RevMan 5.3 according to published guidelines.[27] A random effects meta-regression analysis was conducted to better understand potential sources of heterogeneity of the primary outcome, specifically 1-year overall patient survival. The selection of covariates to include as moderator in the meta-regression model was based on their clinically likelihood to modify the outcome of interest and possible statistically significant different distributions between LDLT and DDLT patients that resulted from the meta-analysis. Meta-regression analysis was conducted using Metafor-package for R studio (version 3.6.3).