OUP user menu

Reply

Padhraig S. Fleming, Ama Johal, Nikolaos Pandis
DOI: http://dx.doi.org/10.1093/ejo/cjt015 548-549 First published online: 5 April 2013

We would like to thank the authors for their interest in our review. We appreciate the comments which highlight some of the difficulties in conducting systematic reviews and meta-analyses.

The decision to conduct quantitative synthesis is often somewhat subjective and opaque. It is unrealistic to expect trials from different settings to be identical in all respects; therefore, discretion is invariably required to assess their similarity. In this instance, we felt laceback use in the upper and lower arches to be comparable as they are applied and act in an identical manner. Furthermore, although different measurement techniques were used in the two studies, both recorded the same outcome: antero-posterior change in incisor position. In view of the overlap of the confidence intervals (CIs), low statistical heterogeneity, allied to what we regarded as low clinical heterogeneity, it was decided that synthesis was reasonable using a random effects model. Furthermore, although only one of the two studies found a significant effect, the direction of the effect in the studies was consistent. Moreover, the range of the CIs did not include genuinely important clinical effects, particularly in view of the potential measurement errors the authors refer to.

Missing data are often problematic in clinical trials; however, if the data are missing at random, the likely consequence is dilution of the effect rather than biased inferences (Carpenter and Kenward, 2008). Adjudication of risk of bias necessitates assumptions and inferences, with varying levels of agreement among assessors (Hartling et al., 2011). In both included studies, loss to follow-up was relatively balanced in both groups; reasons for failure to complete the study were also outlined in participant flow diagrams. Furthermore, even if, as the authors suggest, an unclear risk of bias judgment were given, according to Cochrane guidelines, meta-analysis would still be legitimate.

The authors had concerns that differences in baseline canine angulation between the respective groups may have resulted in biased estimates. Although differences in baseline characteristics can confound the results of a trial, robust randomization procedures were implemented in both included studies. Consequently, baseline differences are less likely to be a problem, and would arise randomly. In fact, the paper by Usmani et al. (2002) reported the following mean canine angulation: 82.6 (9.0) 80.8 (8.0) [right side], 79.8 (10.9) 79.8 (9.3) [left side] for the laceback and the control group, respectively. Given the potential measurement error, such minor differences are likely to be insignificant.

Finally, our conclusions do indicate that ‘on the basis of the available evidence, the use of lacebacks has neither a clinically nor a statistically significant effect on the sagittal position of the incisors and molars during initial orthodontic alignment’. We consider the clinical effect to be of greater importance than statistical significance; our interpretation was made on that basis.

References