Stability and predictability of orthognathic surgery.

In assessing post-treatment stability in a group of treated patients, most of the change usually occurs in just a few of them. For this reason, it is highly misleading to use statistics based on normal distribution to describe posttreatment changes. With a normal distribution, the mean is the most likely indicator of what a patient would experience, and the clinician tends to think of it in just that way. But if essentially no change occurred in three fourths of the patients who underwent a certain type of treatment, and relatively large changes occurred in the one fourth who experienced change, the mean is highly misleading as an expectation of treatment response. It is even more misleading to describe stability in terms of the percentage of treatment change that was retained at some follow-up time, as was done in many early articles on stability after orthognathic surgery. Reporting such percentages implies that the amount of relapse is directly related to the amount of treatment change: the more you changed it, the more it would relapse. In dentofacial patients, that almost never is the case. You simply cannot say that 80% of the amount of typical mandibular advancement will be retained, for instance, because, up to 8–10 mm, posttreatment change (in the few patients who experience it) is relatively independent of the amount of advancement. So how should stability data be reported? The best way is in terms of the percentage of the patients who have changes of a given magnitude. From that perspective, responses can be grouped as: Highly stable—less than a 10% chance of significant posttreatment change Stable—less than a 20% chance of significant post-treatment change and almost no chance of major posttreatment change Stable if modified in a specific way (eg, rigid internal fixation [RIF] after surgery) Problematic: a considerable probability of major posttreatment change In the real world, nothing is 100% successful, and high-risk procedures sometimes are quite successful. The clinician needs to know the odds of long-term stability and predictability with the possible treatment approaches, so that this information can guide the choice of treatment. The goal of this article is to put what we know in that context.