Experimental studies suggest that the regenerating liver provides a “fertile field” for the growth of hepatocellular carcinoma (HCC). However, clinical studies report conflicting results comparing living donor liver transplantation (LDLT) and deceased donor liver transplantation (DDLT) for HCC. Thus, disease-free survival (DFS) and overall survival (OS) were compared after LDLT and DDLT for HCC in a systematic review and meta-analysis. Twelve studies satisfied eligibility criteria for DFS, including 633 LDLT and 1232 DDLT. Twelve studies satisfied eligibility criteria for OS, including 637 LDLT and 1050 DDLT. Altogether, there were 16 unique studies; 1, 2, and 13 of these were rated as high, medium, and low quality, respectively. Studies were heterogeneous, non-randomized, and mostly retrospective. The combined hazard ratio was 1.59 (95% confidence interval [CI]: 1.02–2.49; I2 = 50.07%) for DFS after LDLT vs. DDLT for HCC, and 0.97 (95% CI: 0.73–1.27; I2 = 5.68%) for OS. This analysis provides evidence of lower DFS after LDLT compared with DDLT for HCC. Improved study design and reporting is required in future research to ascribe the observed difference in DFS to study bias or biological risk specifically associated with LDLT.
In the United States, the incidence of hepatocellular carcinoma (HCC) has increased annually by 4.3% since 1992 to the current rate of 5.1 per 100 000, with a one-yr survival rate of only 47% . Liver transplantation is often the optimal curative treatment for HCC . Unlike other treatment modalities, liver transplantation corrects underlying liver dysfunction and removes the diseased tissue that poses a risk for the development of additional future HCC . Unfortunately, the need for donor livers exceeds organ availability in most countries. For example, in the United States, over 18 000 people await deceased donor livers annually, while only approximately 5000 are available . This shortage is likely to worsen; liver donation in the United States has been declining , while HCC incidence has increased .
In the past decade, living donor liver transplantation (LDLT) has received interest as an alternative to deceased donor liver transplantation (DDLT). LDLT alleviates the shortage of donor livers and may allow for the expansion of recipient criteria . However, animal studies suggest that the regenerating liver provides a “fertile field” for the recurrence of HCC after transplantation [7, 8]. Clinical studies report mixed results when comparing LDLT and DDLT: some conclude that LDLT is an inferior “cancer curing” treatment [9-11], while others report no significant differences between the procedures [3, 6, 9-19].
In this study, a systematic review and meta-analysis were performed to assess the current evidence comparing disease-free survival (DFS) and overall survival (OS) after LDLT and DDLT for HCC.
The methods for the systematic review and meta-analysis were defined a priori as outlined below.
Systematic search and data collection
On April 23, 2012, the MEDLINE, Embase, and PubMed databases were searched for the following search terms: “liver transplantation” or “liver transplant,” and “living donor” or “live donor” or “living related,” and “HCC.”
All abstracts were reviewed and assessed by two authors (RG, LS) according to the eligibility criteria in Table 1. The endpoints of DFS and OS were considered separately. Studies that fit the eligibility criteria based on the abstract were obtained, and the manuscripts were reviewed in full (details are found below). The bibliographies of reports meeting the eligibility criteria were examined to identify additional studies.
Table 1. Eligibility criteria for the systematic review and meta-analysis
Studies included in the systematic review
1. Reported original or the most recent results on disease-free survival (DFS) and/or overall survival (OS) after living donor liver transplantation (LDLT) and DDLT for hepatocellular carcinoma in a prospective or retrospective cohort, regardless of the primary focus of the paper
2. Published as articles in a peer-reviewed journal
3. Included a minimum of 5 LDLT and DDLT patients (to exclude case reports)
Studies included in the meta-analysis met the above criteria and
4. Reported sufficient information on DFS and/or OS to calculate unadjusted hazard ratios, and their standard errors, using the algorithm described by Tierney et al. 
Data quality score
A data quality score (DQS) was created to evaluate the quality of evidence in each study, adapted from Sint Nicolaas et al.  (Table 2). Each study was assessed on nine “yes” or “no” questions. The maximum DQS was 12 points. Studies with DQS over 8, 6–8 inclusive, and under 6 were classified as high, medium, and low quality, respectively.
Table 2. Data quality scoring system. Except where noted in brackets, studies were given 1 point for criterion satisfied for living donor liver transplantation (LDLT) and DDLT separately, and 0.5 points for criterion satisfied for LDLT or DDLT, or the whole cohort
1. Prospective study design (1) or a priori analysis (0.5)
2. Same eligibility criteria for LDLT and DDLT (1) or reported as different (0.5)
3. Cohort size exceeded 100 patients
4. Etiology of concurrent liver disease reported
5. Immunosuppressant reported
6. Histological tumor characteristics reported
7. Size and number of tumors reported
8. Time spent on waitlist reported
9. Median follow-up time reported
10. Statistical analysis: multivariable time-to-event model with covariate selection described (3); multivariable time-to-event model without variable selection described (2); stratification on confounders (1)
Two authors (RG, PD) independently entered baseline characteristics, DQS, and outcomes from eligible studies into a Microsoft Excel spreadsheet (Microsoft, Redmond, WA, USA). If estimates were given for multiple time points, the values at transplant were recorded. Where available, the recorded tumor characteristics were obtained from pathological reports. The kappa statistic was calculated to assess inter-rater agreement on the DQS . Any discrepancies on data entry or quality scoring were resolved through consensus with a third author (IM).
Unadjusted hazard ratios (HRs) for DFS and OS comparing LDLT to DDLT were extracted from studies using the algorithm and software from Tierney et al. . Combined HRs were estimated using the random effects model described by DerSimonian and Laird . Combined HRs were estimated separately for Eastern and Western studies in subgroup analysis. Inter-study heterogeneity was assessed using the I2 statistic . Forest plots were used to graphically display the individual study and combined HR. Funnel plots and Egger's regression test were used to assess publication bias . Analysis was performed using the “metafor” package  in R . Statistical significance was set at p < 0.05.
Systematic literature search and data entry
Nine hundred and fifty-six studies were identified in Embase and MEDLINE and 555 in PubMed, 792 of which were duplicates. Sixteen unique studies satisfying the eligibility criteria the systematic review described in Table 1 were selected [3, 6, 9-19, 28-30]. Of these, 12 studies satisfied the eligibility criteria for DFS, including 633 LDLT and 1232 DDLT [3, 6, 9-11, 14, 16-19, 28, 30], and 12 studies satisfied the eligibility criteria for OS, including 637 LDLT and 1050 DDLT [3, 6, 9-19, 28-30]. The most recent study from the A2ALL group omitted information required for the meta-analysis for OS; therefore, earlier studies from these centers were used in the meta-analysis [9, 29]. Searching the bibliographies of eligible studies did not identify additional studies. Colleagues translated foreign language abstracts; however, none met the eligibility criteria.
Table 3 displays select study characteristics. Studies varied widely in their baseline characteristics, and certain characteristics were infrequently reported.
Table 3. Characteristics of the studies included in the systematic review
Median months on waitlist (range)
Median months of follow-up (range)
% outside Milan criteria
DDLT, deceased donor liver transplantation; LDLT, living donor liver transplantation.
Values given are in terms of mean ± standard deviation.
Concurrent liver disease was consistent with regional prevalence of hepatitis C and B viruses . Studies reporting neoadjuvant therapy used different combinations of radiofrequency ablation, hepatic resection, transarterial chemoembolization, and/or percutaneous ethanol injection, and studies reporting immunosuppression used mainly calcineurin inhibitors and corticosteroids.
The median time from listing to transplant was significantly shorter for LDLT than DDLT in 6 of 7 studies that reported waitlist time in both cohorts. Median follow-up time was significantly shorter for LDLT compared with DDLT in 3 of 5 studies that reported follow-up in both cohorts.
One of 9 of the studies reporting the proportion of patients exceeding the Milan criteria at the time of transplant found that significantly more LDLT exceeded the criteria; the other 8 reported no significant differences. The reporting of other measures of disease severity in the LDLT and DDLT cohorts was inconsistent across studies: AFP levels were reported in 8; CTP classification in 3; MELD score in 7; number of tumor nodules in 9; tumor diameter in 10; tumor stage in 5; degree of differentiation in 4; microvascular invasion in 5; and macrovascular invasion in 3. However, few significant differences existed between LDLT and DDLT in studies for reported measures of disease severity: in Fisher et al., LDLT had a significantly higher proportion of patients with T3 or T4 tumor stage at transplant and higher AFP ; in Hwang et al., LDLT had a significantly worse CTP classification ; and in Lo et al. , LDLT had a significantly better CTP classification.
Data quality score
The 16 unique studies were evaluated according to the DQS described in Table 2; 1, 2, and 13 were scored as high, medium, and low quality, respectively (Figs. 1 and 2). The kappa statistic from DQS was 0.85, due to two errors that were corrected. No study reported prospective study design or specified analysis a priori. Four studies explicitly stated that the same recipient eligibility criteria were applied to LDLT and DDLT. LDLT and DDLT each numbered <100 in all studies; however, LDLT and DDLT combined exceed 100 patients in 4 studies. 1 study reported on concurrent liver disease, immunosuppressant protocol, histological tumor characteristics, size and number of tumors, time spent on the waitlist, and median follow-up time separately for LDLT and DDLT. Only two and three studies performed any form of statistical adjustment for confounding factors for DFS and OS, respectively.
Disease-free survival: study results and meta-analysis
DFS was significantly shorter after LDLT in three of 10 studies that reported the log-rank test, and significantly longer in 0 of 10 (Fig. 1). Three-yr DFS ranged from 42% to 95.5% after LDLT and from 41% to 100% after DDLT.
DFS estimates after adjusting for confounding were reported in two studies [9, 30]. In Fisher et al., time to recurrence remained significantly shorter after LDLT compared with DDLT after stratification by tumor stage . For the combined endpoint of death or recurrence, the HR for DFS after LDLT vs. DDLT was 0.82 (95% confidence interval [CI]: 0.38–1.72) after adjusting for transplant year, AFP, recipient age, and transplant center . Sandhu et al.  found no significant differences in recurrence rates between LDLT and DDLT after adjusting for AFP, number and size of tumors, the proportion of patients outside the Milan criteria, and micro- and macrovascular invasion (HR = 0.66, 95% CI: 0.24–1.84).
The HR for DFS after LDLT vs. DDLT was 1.59 (95% CI: 1.02–2.49; p = 0.041), combining the 8 cohorts with sufficient information for meta-analysis (Fig. 1). Moderate statistical inter-study heterogeneity was present (I2 = 50.07%), likely resulting from variations in patient characteristics and treatment protocols between studies. Publication bias was not detected when assessing the funnel plot or with Egger's regression test (p = 0.23) (data not shown). Studies from Eastern [3, 10, 17] and Western countries [6, 9, 11, 12, 14-16, 19, 30] had similar HR for DFS: 1.56 (95%CI: 0.85–2.86; p = 0.154; I2 = 43.94%); and 1.62 (95% CI: 0.79–3.34; p = 0.191; I2 = 61.73%), respectively.
The mean time from transplant to HCC recurrence in these studies ranged from 4.6 to 38 months, and the median time from recurrence to death ranged from 6 to 30.6 months.
Overall survival: results and meta-analysis
OS was significantly longer after LDLT compared with DDLT in 3 of 10 cohorts reporting log-rank tests, and significantly shorter in 0/10 (Fig. 2). Three-yr survival ranged from 53% to 80% after LDLT, and 50% to 90% after DDLT.
OS estimates after adjusting for confounding were reported in two studies. In Sandhu et al., the HR for OS after LDLT vs. DDLT was 0.64 (0.23–1.79) after adjusting for age, AFP, number and size of tumor nodules, the proportion of patients exceeding the Milan criteria, and micro- and macrovascular invasion . Berg et al. reported the HR for OS to be 2.17 (p = 0.17) in candidates for transplant MELD scores under 15 at evaluation, and 1.1 (p = 0.91) for candidates with MELD scored over 15 after adjusting for MELD at transplant, age at transplant, HCV infection, and cholestatic liver disease .
The HR for OS after LDLT vs. DDLT was 0.97 (95% CI: 0.73–1.27; p = 0.808), combining the 8 cohorts with sufficient information for meta-analysis. Statistical inter-study heterogeneity was low (I2 = 5.68%). Publication bias was not detected when assessing the funnel plot or with the Eggerton regression test (p = 0.09) (data not shown). Eastern studies [3, 13, 17] had a lower HR for OS compared with Western studies [6, 9, 11, 19, 30]: 0.74 (95% CI: 0.50-1.10; p = 0.141; I2 = 6.21%) and 1.23 (95% CI: 0.85–1.78; p = 0.269; I2 = 0.00%), respectively.
The present analysis suggests that DFS is worse after LDLT compared with DDLT for HCC. As a decreased DFS implies higher rates of post-transplantation interventions and may eventually translate to decreased OS, we recommend that the increased risk of recurrence be discussed with all potential donors and recipients who are considering LDLT for HCC. More research is needed to determine whether the observed differences are due to study biases or to truly inferior biological outcomes after LDLT. More thorough reporting and analysis in future clinical studies would facilitate this research.
Individual studies of hepatocellular cancer recurrence rates after DDLT and LDLT have yielded conflicting results. This study examined this conflict using a systematic review and meta-analyses, which allows one to aggregate evidence and potentially minimize bias . This approach addresses three main shortcomings in the previous literature. First, combining all published data increases the power to detect a significant difference in DFS or OS. Second, simultaneously analyzing the results of all studies identified through a systematic search mitigates study- and center-specific biases. Finally, for the first time, DFS and OS after LDLT and DDLT for HCC were analyzed using a systematic approach defined a priori.
This systematic review identified significant gaps in the quality of previous clinical reports using a DQS. Most studies had major deficiencies, including poor reporting of baseline patient characteristics and inadequate statistical approaches. To improve the quality of the evidence comparing LDLT and DDLT for HCC, future reports should follow the standardized reporting guidelines endorsed by the STROBE consortium  and include all attributes assessed in the DQS instrument adapted for this review.
Although LDLT was associated with significantly decreased DFS in these studies, there was no difference observed in OS. However, the median follow-up ranged from 3 months to 58 months, the mean time from transplant to HCC recurrence ranged from 4.6 to 38 months, and the median time from recurrence to death ranged from 6 to 30.6 months. Hence, the follow-up periods in these studies are probably too brief to detect a major impact on survival resulting from differences in disease recurrence.
The search strategy and eligibility criteria used in this systematic review and meta-analysis included cohort studies of both LDLT and DDLT for HCC. Reports that included LDLT-only and DDLT-only groups transplanted for HCC were excluded from this analysis to minimize the risks of confounding biases caused by differences in patient selection over time, pre- and post-transplant protocols, the experience of the surgical teams, and the duration of follow-up. Despite this restriction, our search strategy and eligibility criteria captured studies from the diverse geographical regions. Such diversity is important because underlying disease etiology, donor and recipient selection procedures, and cultural and ethical conditions differ by region, particularly between Eastern and Western nations [34, 35]. Nonetheless, DFS was worse after LDLT compared with DDLT in both Eastern and Western regions; however, DFS was significantly worse only when Eastern and Western regions were combined, demonstrating the power of meta-analysis. OS after LDLT vs. DDLT was better in studies from Eastern centers compared with studies from Western centers, although the difference between LDLT and DDLT was non-significant in both subgroups. This may reflect differences between these regions in patient selection or case-volumes effects on surgical outcomes.
As LDLT alleviates the growing donor liver shortage and might allow for the expansion of recipient eligibility criteria, more research is needed to determine the causes of the shortened DFS after LDLT vs. DDLT. Proposed mechanisms by which LDLT for HCC might confer an increased risk of recurrence over DDLT include (i) the release of hepatotrophic cytokines and increased vascular inflow associated with hepatic regeneration may enhance the growth of residual HCC cells, which is evidenced by animal and in vitro human models [7, 8, 14]; (ii) sparing the inferior vena cava during LDLT may compromise complete tumor removal [9, 14]; and (iii) greater manipulation of the liver during LDLT may promote HCC dissemination .
Alternatively, the shorter DFS observed after LDLT might be caused by study biases rather than true biological differences in the surgical treatments. For instance, patients with aggressive tumor biology may have been preferentially allocated to LDLT. The risk of bias caused by differences in tumor biology is difficult to evaluate, because the patient characteristics were poorly reported. Similarly, the briefer wait-list time observed with LDLT could limit the detection of particularly aggressive tumors . The possibility that the worse DFS observed after LDLT might be caused by biases underscores the necessity for thorough reporting of patient characteristics and adequate statistical adjustment for confounding.
In summary, this systematic review and meta-analysis suggested decreased DFS after LDLT compared with DDLT for HCC. However, given that existing studies are heterogeneous and of low quality, further research is needed to determine whether this is due to study biases or to biological differences.
All authors contributed to study design. R.C.G., L.S., and P.R.D. collected the data. R.C.G. analyzed the data and wrote the draft manuscript. All authors reviewed and contributed to the final manuscript.