Eight studies were included in the review (n=1,188 patients). The number of patients per treatment group ranged from 18 to 155.
All studies were reported as scoring 3 points of the Jadad scale. The quality table reported that all were described as randomised and all used appropriate methods of sequence generation. None were double-blinded and none specifically described allocation concealment.
There was no statistically significant difference between hepatectomy and radiofrequency ablation in the complication rate (three studies) or death rate (three studies provided data).
Survival rates were significantly lower in the radiofrequency ablation group at three years (OR 1.66, 95% CI 1.21 to 2.27; six studies), but there was no significant difference in survival rates between radiofrequency ablation and hepatectomy at one year (seven studies), two years (two studies) and five years (six studies).
Recurrence-free survival rates were significantly lower in radiofrequency ablation groups at one year (OR 1.54, 95% CI 1.07 to 1.21; five studies), three years (OR 2.02, 95% CI 1.48 to 2.75; five studies) and five years (OR 1.52, 95% CI 1.12 to 2.06; six studies).
Radiofrequency ablation was associated with a significantly higher rate of recurrence at the previous site (OR 0.28, 95% CI 0.12 to 0.65; four studies) and a significantly lower rate of recurrence in new areas (OR 1.97, 95% CI 1.29 to 2.99; four studies). There was no statistically significant difference between hepatectomy and radiofrequency ablation for recurrence in extra-hepatic areas (two studies).
The authors stated that there was no significant heterogeneity was found for any of these analyses. Forest plots showed that I2 values of 50% or more for analyses of three-year survival plus one- and three- and five-year recurrence-free survival. Studies of five-year recurrence-free surgical showed different directions of treatment effect.
The authors stated that the funnel plot of survival at five years showed symmetry ‘indicating no serious publication bias’. The funnel plot appeared asymmetrical on inspection.