Twenty-three studies (2,051 patients) were included in the review. Adequate randomisation was reported in 17 studies, and 11 studies reported satisfactory allocation concealment measures. Only one study reported adequate blinding and eleven studies reported sample size calculations and defined primary outcomes. Five studies reported drop-outs and of these, two studies reported using intention-to-treat analyses. Sixteen studies appeared to be free from other sources of bias.
Random-effects models were used for operation time, blood loss, hospital stay and return to normal activities because of significant statistical heterogeneity.
Statistically significant benefits were observed with LAVH compared with abdominal hysterectomy with reductions in blood loss (WMD -47.92ml, 95% CI -77.79 to -18.06; Ι²=91%; 13 studies), hospital stay (WMD -2.11 days, 95% CI -2.63 to -1.59 days; Ι²=96%; 13 studies), smaller haemoglobin drop (WMD -0.52 g/100ml, 95% CI -0.73 to -1.31g/100 ml; Ι²= 0%; four studies), quicker time to normal activity (WMD -13.32 days, 95% CI -16.67 to -9.88 days, Ι²= 71%; six studies) fewer minor complications (OR 0.50, 95% CI 0.36 to 0.70; Ι²=14%; 17 studies) fewer overall complications (OR 0.60, 95% CI 0.44 to 0.81, Ι²= 9%, 18 studies) and less post-operative pain on days one to three post-surgery.
The use of LAVH was associated with significantly longer operation time (WMD 13.62 minutes, 95% CI 4.60 to 22.65 minutes, Ι²= 96%, 15 studies) and more major complications (OR 2.54, 95% CI 1.13 to 5.70, Ι²=0%, eight studies) compared with abdominal hysterectomy.
Meta-regression analyses showed that power calculations and adequate sequence generation were potential sources of heterogeneity for outcomes of length of hospital stay. No single study had an undue influence of overall estimates of mean differences. Sensitivity analyses showed similar results were obtained with fixed-effect and random-effects models except LAVH was associated with shorter operation times, greater blood loss, shorter hospital stay and a slightly slower return to normal activity when the results were analysed using a fixed-effect model.
The results of the Egger's test showed no evidence of publication bias.