Four randomised controlled trials (RCTs; n=152) and 6 retrospective comparative studies (n=756) were included in the review.
In only 4 of the 10 studies were the patient characteristics deemed similar between groups at baseline in terms of risk of death during surgery. Adequate adjustment for potential confounding was made in 6 studies. Attrition was accounted for, and follow-up times were consistent between groups in all studies.
There was no statistically significant difference in mortality between patients receiving the maze procedure combined with mitral valve surgery and those receiving mitral valve surgery alone (8.4% versus 5.8%; 4 RCTs).
The stroke rate was statistically significantly reduced in the maze group compared with the control group (0% versus 5.8%; effect size 0.44, 95% confidence interval, CI: 0.12, 0.77, p=0.008; 4 RCTs). The sensitivity analysis in which one study was removed from the analysis reduced the difference to a non significant level.
Sinus rhythm restoration was statistically significantly higher in the maze group compared with the control group (80.7% versus 17.3%; effect size 1.39, 95% CI: 1.06, 1.71, p<0.000001; 4 RCTs).
The need for pacemaker following surgery was statistically significantly higher in the maze group compared with the control group (3.9% versus 1.5%; effect size -0.16, 95% CI: -0.29, -0.02, p=0.02; 10 studies). The sensitivity analysis using Hedges d statistic rather than Cohen's h reduced the difference to a non significant level.
The post-operative bleeding rate of maze with surgical incisions was statistically significantly higher than that for mitral valve surgery alone (4.3% versus 0%; effect size -0.41, 95% CI: -0.71, -0.11, p=0.007; 3 studies). The sensitivity analysis in which one study was removed from the analysis reduced the difference to a non significant level. The bleeding rate did not differ between radiofrequency maze and mitral valve surgery alone (1.9% versus 2.9%; 5 studies).
None of the meta-analyses showed statistically significant heterogeneity, and for all outcomes the results of fixed-effect and random effects meta-analyses did not differ.