Comparison of outcomes of transplantation and resection in patients with early hepatocellular carcinoma: a meta-analysis

Abstract

Objectives: Surgical decision making for patients with early hepatocellular carcinoma (HCC) and well-compensated cirrhosis remains controversial. The aim of the current study was to conduct a meta-analysis of published reports to compare survival outcomes after transplantation and resection, respectively, in patients with early HCC [i.e. HCC falling within the Milan Criteria (a solitary lesion measuring ≤5 cm or fewer than three lesions with a largest diameter of ≤3 cm, and absence of macroscopic vascular invasion or extrahepatic disease)] and well-compensated cirrhosis.

Methods: A total of 990 abstracts were identified through a PubMed-based search. Ten articles comparing transplantation and resection in patients with early HCC were included in the meta-analysis. Meta-analysis was performed using stata 9.2 statistical software.

Results: Outcomes were analysed for a total of 1763 patients with early HCC. The 5-year overall survival (OS) for all patients was 58% (transplantation: 63%; resection: 53%). Meta-analysis of all 10 studies revealed a survival advantage for transplantation [odds ratio (OR) 0.581, 95% confidence interval (CI) 0.359-0.939; P = 0.027]. Analysis of only those reports that utilized an 'intention-to-treat' strategy failed to demonstrate a survival advantage for either treatment approach (OR 0.600, 95% CI 0.291-1.237; P = 0.166).

Conclusions: The current study demonstrates a favourable outcome in patients with early HCC treated by either transplantation or resection. Although transplantation was noted to have a survival advantage in some settings, resection continues to be a viable treatment approach.

Figure 1

Flow diagram showing search strategy along with the selection and screening process for the eligible studies. HCC, hepatocellular carcinoma

Figure 2

Meta-analysis of 5-year overall survival utilizing data from 10 studies comparing outcomes after transplantation and resection, respectively, in early hepatocellular carcinoma (i.e. within the Milan Criteria). Odds ratios (ORs) for overall survival in the transplantation and resection subgroups were calculated using the random-effects model. The diamond represents the overall effect; squares represent the effects for individual studies; bars indicate 95% confidence intervals (CIs)

Figure 3

Funnel plot for the 10 studies comparing outcomes after transplantation and resection, respectively, in patients with early hepatocellular carcinoma

Figure 4

Meta-analysis of 5-year overall survival utilizing data from six studies comparing outcomes after transplantation and resection, respectively, in early hepatocellular carcinoma (i.e. within the Milan Criteria) in patients with well-compensated cirrhosis. Odds ratios (ORs) for 5-year overall survival in the transplantation and resection subgroups were calculated using the random-effects model. The diamond represents the overall effect; squares represent the effects for individual studies; bars indicate 95% confidence intervals (CIs)

Figure 5

Meta-analysis of 5-year overall survival utilizing data from six studies comparing outcomes after transplantation and resection, respectively, in early hepatocellular carcinoma (i.e. within the Milan Criteria), utilizing an intention-to-treat analysis. Odds ratios (ORs) for 5-year overall survival in the transplantation and resection subgroups were calculated using the random-effects model. The diamond represents the overall effect; squares represent the effects for individual studies; bars indicate 95% confidence intervals (CIs)

Figure 6

Meta-analysis of 5-year overall survival utilizing data from three studies comparing outcomes after transplantation and resection, respectively, in early hepatocellular carcinoma (i.e. within the Milan Criteria) in patients with well-compensated cirrhosis, using an intention-to-treat strategy. Odds ratios (ORs) for 5-year overall survival in the transplantation and resection subgroups were calculated using the random-effects model. The diamond represents the overall effect; squares represent the effects for individual studies; bars indicate 95% confidence intervals (CIs)