We read the published systematic review by Mattos et al. with interest. The authors used the methods of systematic review and meta-analysis to explore the changes in the airway of adult patients who underwent orthognathic surgery to correct anteroposterior osseous discrepancies. This article provided much clinical evidence for clinicians and we think it will play an important role in clinical decision making.

The authors also used meta-analysis to compare the preoperative and postoperative values (derived from one group of patients in a study). This is a method we have seldom seen and it provides some new ideas for us to further practice evidence-based dentistry, but we have some opinions on the method that we think might lead to more accurate outcomes.

First, the authors directly input the mean and SD values of the pre- and postoperative data into Revman to obtain the forest plot; and we do not think that such a method could reflect the actual outcome of the original studies accurately. As in the original studies, the pre- and postoperative data were compared in a paired way (for example, the original authors might use a paired Student’s t -test). When we input these data directly into Revman they will be treated as group data and the 95% CIs of these outcomes will be wider than they actually were (which is why we could use Revman to analyze randomized controlled trials (RCTs), as RCTs could not be paired). The solution to such a problem is quite simple. The Cochrane Handbook introduced a way of dealing with paired outcomes, which is to calculate or assume the correlation coefficient (Corr), and use Corr to calculate the SD of the mean difference. The Corr could be assumed as 0.5. The stability of the outcomes should be detected by sensitivity analysis by assuming Corr as 0.7 and 0.3.

Second, the authors mentioned that a fixed-effects model would be used if I ² < 75%. The Handbook suggests that a fixed-effects model should be used when I ² < 50%, not 75%; and for I ² = 75%, it is the threshold whether to conduct a meta-analysis.

Overall, we think these suggestions might make the outcomes more accurate, but we still hope that we could discuss these with the authors as the method, which compares the pre- and postoperative data in the meta-analysis, is quite new and still need some improvement.