In this issue of CHEST (see page 1184), Vezina et al1 evaluate the long-term risks and efficacy (>2 years) of two different bibloc devices (ie, traction and compression) based on regular medical visits, questionnaires, cephalometrics, and level 1 and 3 polysomnograms in >150 subjects who were moderately overweight but nonobese. Unfortunately, only 40% of the subjects fitted with an oral device were seen in follow-up visits, and only 26.5% (43 of 162) of the subjects had follow-up polysomnograms. Although Vezina et al1 did not design the study to compare the efficacy of the devices, they did report that 52% and 61% of the subjects outfitted with a compression appliance and a traction appliance, respectively, demonstrated partial efficacy, with a mean improvement in the AHI from around 30 events per hour at baseline to around 14 events per hour with treatment. They reported pain, especially at the beginning of treatment, as the primary side effect, which the authors deemed clinically irrelevant. The study closed with a rather small sample size, with a large number of nonparticipating subjects. Vezina et al1 did not randomize the type of devices, but instead preferentially prescribed traction-based devices to lean female subjects and compression-based appliances, which have more robust attachments, to male subjects and to subjects with symptoms of bruxism. Therefore, it is difficult to draw conclusions with regard to long-term changes in dental occlusion based on appliance type. There has been a shift over the years from monobloc to bibloc appliances to minimize complications and improve efficacy. However, the type of bibloc device selected is likely as important as selecting good candidates for appliance therapy. Our general lack of understanding of the long-term risks and physiologic actions of each device makes it difficult to select the most appropriate device. This work suggests that proper appliance selection might minimize pain and ultimately improve compliance. There is a need for a standard algorithm for appliance selection.