Use of facial features to predict obstructive sleep apnea presence and severity

A recent study by Vidigal et al. [1] explored the link between the presence and severity of obstructive sleep apnea (OSA) and measurements made on craniofacial and intraoral photographs in a large adult population. OSA is a serious health issue, and there is a significant interest in identifying how certain craniofacial characteristics could be linked to OSA development or are OSA predictors. Ultimately, one could hypothesize that modifying the relevant craniofacial features through orthodontic treatments could help manage or prevent OSA. This audacious goal needs to start with a clear understanding of OSA pathophysiology and its link to craniofacial growth and development. Hence, this study is timely and asks an important question. This study has several notable strengths. The authors used a large, population-based sample, employed accepted reference methods for OSA diagnosis, and a standardized photographic methodology. They studied patients from a general population, as well as a subset of outpatients from a sleep institute. Notably, these patients belong to a large, long-term collaborative project—Sleep Apnea Global Interdisciplinary Consortium. They have also undertaken a considerably comprehensive statistical assessment of their data, and their findings have been reasonably presented. Nonetheless, in our opinion, the results appear to be over-interpreted, and the conclusions drawn are overreaching and not fully supported by the findings. In sum, only minor, clinically insignificant differences between groups are noted (Table 2), and one can argue that based on the reported data, the severity of OSA does not consistently correlate with tongue size and position or other characteristics of the face and lower jaw. The presented data is aligned with a recent related meta-analysis [2], where the portrayed associations were not statistically important. Specifically, two important notions support our interpretations stated above. First, when we focus on the results that more comprehensively adjust for age, sex, and body mass index, no significant differences are seen between apnea-hypopnea index (AHI) groups with craniofacial characteristics in the general population (Table 4). This suggests that purely anatomical assessments are poor predictors of the presence or severity of OSA, especially in a general non-symptomatic population. Second, of approximately 65 individual facial or intraoral measurements reported in this study, only a handful showed statistical significance when assessing between the general and clinical populations. Hence, most variables assessed were not significantly different between populations or between individuals with high and low AHI (Table 2). Our interpretation considers another limitation of this article: a lack of reporting on measurement error. While the photogrammetry setup was standardized, the actual identification of landmarks by the operators, upon which the measurements were based, can vary. Since most of the linear differences found are small (<1 mm), if measurement errors are relatively large, they can certainly overshadow statistically significant findings. Landmark identification errors can impact angular and area measurements even more since three and four landmarks are used, respectively, compared to linear measurements, which use only two landmarks. Intraorally, the only finding significantly associated with OSA was a minor difference in the curvature of the tongue. Importantly, this measurement was not different in the general population group and was only smaller by 0.5 mm in the OSA clinical group compared to the clinical non-OSA group. It is also important to clarify that different authors use the term tongue curvature differently. Tongue curvature is often described as the posterior region based on a lateral cephalogram or MRI imaging. The diminished curvature of the anterior velopharyngeal wall is related to a less perpendicular airway, with a better airway dynamic [3]. The static size, composition, and shape are part of the passive features of the tongue and airway. To understand upper airway collapse, we must look beyond the active components involving muscle activity produced by reflex and central drives and tissue deformation under breathing pressures, which collectively contribute to (or not) a patent airway [4]. The tongue curvature, in the current paper, is the distance in the circumference of the tongue when the tongue is positioned outside of the mouth. Since all the other measurements of the tongue in the lateral view, such as length, width, and volume, are not significant, it is unclear if this measurement has any clinical or physiological meaning. We must not forget that whenever multiple individual variables are being tested, there is always a risk of a type I error, which cannot be ruled out here. Additionally, a binary assessment -logistic regression modeling—of a multi-factorial disease, i.e., OSA, to identify predictors introduces threshold bias and loss of information. Converting continuous data into binary data is highly dependable on the threshold level. In the assessment of tongue position, the authors also failed to emphasize that the biggest and strongest difference was the Mallampatti score, which, despite being subjective, had the best associations with the presence and severity of OSA. The typical facial morphology postulated to be associated with OSA is a long face, retruded mandible, with significant facial convexity according to a recent research synthesis [5], which contrasts the research synthesis results [2] mentioned above. Surprisingly, mandibular width is the only variable that increased OSA cases in this study, regardless of the population studied. When considered frontally, these characteristics are associated with a class III malocclusion (a prominent mandible, not a retruded one). This does not match the above-stated expected facial characteristics associated with OSA and is, in fact, contradictory to it. The author’s interpretation of this finding is, in our opinion, an example of some overreaching conclusions made. They suggest that “Collectively, these phenotypic findings could reflect the combined effects of excess facial fat/soft tissue volume and mandibular deficiency, suggesting a narrow pattern or a wider face that contributes to the presence of moderate or severe apnea” [1]. How would an increased mandibular width, wider face, and increased mandibular triangle area relate to a smaller bony width of the face? These are contrasting morphological characteristics and the thesis provided here, fails to clarify how these facial morphological findings can be associated with OSA. Indeed, it emphasizes that OSA is a complex clinical phenotype that does not abide by the one-size-fits-all view held by many who suggest that everyone should have a wide maxillary and mandibular dentition to manage or prevent OSA. To be clear (and fortunately), the authors did not suggest this, but it is an important point to emphasize. The discussion could have been presented more simply to reflect that no individual variable per se is likely to be highly predictive of the complexity of OSA phenotyping. In summary, while the traditional clinical phenotype for OSA is presented (male, old, obese, high ODI, and low SpO2), that is not the case for the expected craniofacial features (long narrow faces with mandibular retrusion and constricted maxillary dentition). It is imperative to recognize and communicate that statistical associations or correlations do not amount to causation. This study’s directionality of associations and correlations is not demonstrated based on the provided data. The overarching question is whether OSA development affects craniofacial characteristics or whether these altered characteristics facilitate OSA development, and the current article’s findings do not support any such association, let alone hint at causation. Using photogrammetry to screen large populations is important [6], and the authors’ undertaking of a large-scale, comprehensive study is commendable. Still, the data reported should be carefully considered on its own merits. Despite the known bias of the publication of positive findings, this report could be considered a negative study. Our answer to the study’s title, “Can intraoral and facial photos predict obstructive sleep apnea in the general and clinical population?” based upon an unbiased assessment of the data, is “NO.” The current results highlight the complex phenotypes of OSA cases and the relatively minor importance of craniofacial extraoral and intraoral features in the big scheme. There could be a time when a relative subgroup of OSA is identified where the craniofacial features are a major component. Still, attempts to generalize such associations have not been methodologically strong. The editorial letter was drafted in Edmonton, AB, Canada; Vancouver, BC, Canada; Sammamish, WA, USA. Financial disclosure: None declared. Nonfinancial disclosure: None declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.