Eight RCTs were included in meta-analyses (n=892). Sample size ranged from 26 to 281. Randomisation and allocation concealment were adequate in most of the RCTs. It appeared that intention-to-treat analyses were adequate in all trials. Loss to follow-up ranged from 8% to 33%.
For Trial A neurocognitive test, significant improvements in neurocognitive function were seen in patients who underwent off-pump CABG compared with those underwent on-pump CABG at less than three-month follow-up (WMD -2.62, 95% CI -4.79 to -0.44; six RCTs) and at greater than six-month follow-up (WMD -4.96, 95% CI -7.35 to -2.57; five RCTs).
For Rey Auditory Verbal Learning test, Grooved Pegboard test, Trail B test and Digit Symbol test, there were no significant differences in neurocognitive outcomes between the two groups at less than three-month follow-up and greater than six-month follow-up.
Significant heterogeneity was observed in the outcome at less than three-month follow-up for Grooved Pegboard test (I2=68.5%), Trail B test (I2=70%) and Digit Symbol test (I2=70.1%) and the outcome at greater than six-month follow-up for Rey Auditory Verbal Learning test (I2=79.8%). The authors reported that the funnel plot (not presented) showed no significant publication bias.