Craniofacial risk factors for obstructive sleep apnea—systematic review and meta‐analysis

Obstructive sleep apnea (OSA) is caused by temporary partial or complete constriction of the upper airway during sleep which leads to reduced blood oxygen and cardiovascular risks. Main symptoms vary between adults and children leading to misdiagnosis or delayed patient identification. To improve early diagnosis, lateral cephalograms can provide craniofacial measurements associated with a higher risk of OSA. In order to identify the most relevant craniofacial measurements, a systematic literature review with meta‐analysis was conducted combining the terms ‘orthodontic*’, ‘craniofacial’, ‘cephalometr*’, ‘cephalogram’, ‘OSA*’, ‘UARS’, ‘SDB’, ‘sleep disordered breathing’, ‘sleep apnea’ and ‘sleep apnoea’. Of 3016 publications, 19 were included in the systematic review and meta‐analysis, 15 with adult patients and four with children. A total of 16 measurements (six angles, 10 distances) were compared, nine showed a possible influence in patients with OSA compared to controls: NSBa angle (−0.28°), ANB angle (+0.33°), ML‐NSL angle (+0.34°), Me‐Go‐Ar angle (+0.33°), SN distance (−0.70 mm), N–ANS distance (−0.36 mm), MP–H distance (+1.18 mm), uvula length (+1.07 mm) and thickness (+0.96 mm). Posterior airway measurements were not sufficiently described or comparably measured to be statistically analysed. There is some evidence for altered craniofacial anatomy in patients with OSA compared to controls. Lateral cephalograms should be screened for these aspects routinely to improve early diagnosis of OSA and craniofacial orthopaedics should complement the interdisciplinary treatment plan for young patients with OSA.


Summary
Obstructive sleep apnea (OSA) is caused by temporary partial or complete constriction of the upper airway during sleep which leads to reduced blood oxygen and cardiovascular risks.Main symptoms vary between adults and children leading to misdiagnosis or delayed patient identification.To improve early diagnosis, lateral cephalograms can provide craniofacial measurements associated with a higher risk of OSA.In order to identify the most relevant craniofacial measurements, a systematic literature review with meta-analysis was conducted combining the terms 'orthodontic*', 'craniofacial', 'cephalometr*', 'cephalogram', 'OSA*', 'UARS', 'SDB', 'sleep disordered breathing', 'sleep apnea' and 'sleep apnoea'.Of 3016 publications, 19 were included in the systematic review and meta-analysis, 15 with adult patients and four with children.A total of 16 measurements (six angles, 10 distances) were compared, nine showed a possible influence in patients with OSA compared to controls: NSBa angle (À0.28 ), ANB angle (+0.33 ), ML-NSL angle (+0.34 ), Me-Go-Ar angle (+0.33 ), SN distance (À0.70 mm), N-ANS distance (À0.36 mm), MP-H distance (+1.18 mm), uvula length (+1.07 mm) and thickness (+0.96 mm).Posterior airway measurements were not sufficiently described or comparably measured to be statistically analysed.
There is some evidence for altered craniofacial anatomy in patients with OSA compared to controls.Lateral cephalograms should be screened for these aspects routinely to improve early diagnosis of OSA and craniofacial orthopaedics should complement the interdisciplinary treatment plan for young patients with OSA.

| INTRODUCTION
More than 930 million people worldwide experience obstructive sleep apnea (OSA; Benjafield et al., 2019), which is characterised by repetitive partial or total constriction of the upper airway during sleep and leads to reduced blood oxygen and impaired sleep quality (Juliano et al., 2009;Yadav et al., 2014).According to the American Academy of Sleep Medicine (AASM) guidelines (Kapur et al., 2017), OSA can be diagnosed at an Apnea-Hypopnea Index (AHI) cut-off of ≥5 events/h combined with clinical symptoms.Several studies define the absence of OSA with an AHI of < 10 and/or the diagnosis of OSA at an AHI of ≥ 10 events/h.In children >2 events/h is usually classified as OSA (Katyal, Kennedy, et al., 2013).Symptoms in adults are snoring, awakening, daytime sleepiness and microsleep.Because of possible severe medical consequences such as cardiovascular diseases, depression and reduced physical performance, early diagnosis and treatment is sought.Children with OSA show slightly different symptoms, such as behavioural problems, and are therefore sometimes misdiagnosed as attention deficit disorder (Rosen, 2000).Known risk factors include obesity, alcohol, sleep medication and patient age, but there are as well indicators for relevant craniofacial anomalies that increase the risk of OSA or its severity (Sauer et al., 2012).There are already some systematic reviews on this topic (Agha & Johal, 2017;Armalaite & Lopatiene, 2016;Fagundes et al., 2022;Flores-Mir et al., 2013;Katyal, Pamula, et al., 2013;Miles et al., 1996;Neelapu et al., 2017), but with relatively low numbers of studies included in the meta-analysis for each item.(Broadbent, 1931;Hofrath, 1931).Various authors have contributed to the improvement and refinement of this diagnostic technique leading to a huge number of different evaluation and interpretation approaches (Björk, 1947(Björk, , 1950;;Brodie, 1941;Coben, 1955;Graber, 1952;Johnson, 1950;Ricketts, 1957;Steiner, 1953;Tweed, 1946).Therefore, measurement routines are not consistently used in all countries today, which impedes the comparability of norms as well as individual values and makes scientific research and comparison in this field very difficult.
To identify diagnostically conclusive findings concerning OSA, several research groups conducted clinical trials on lateral cephalograms of patients with OSA and control groups.
The aim of this systematic review was the identification of possibly altered cephalometric values in patients with OSA compared to healthy controls.As a preliminary question, the terminology in cephalometric measurements should be assessed to aim for a consistent terminology.

| METHODS
The systematic review and meta-analysis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher et al., 2015;Shamseer et al., 2014) and was registered with the International Prospective Register of Systematic Reviews (PROSPERO; identifier: CRD42019118288).Additionally, the reference lists of eligible studies were screened for relevant publications.

| Study selection
Inclusion criteria were language (English, German, French), prospective or retrospective clinical studies with control groups, diagnosis of OSA via polygraphy or polysomnography (PSG) with AHI/Apnea Index (AI)/Respiratory Disturbance Index (RDI) ≥ 10 events/h patient groups, lateral cephalometric evaluation of patients and controls with sufficient description of data acquisition and data provided as mean ± standard deviation (SD).Because of inconsistent definition of healthy controls in terms of OSA, a stricter threshold was used in the inclusion criteria of this study compared to the AASM guidelines (Kapur et al., 2017): the threshold was defined as ≥ 10 events/h to ensure that only patients with OSA were included and to differentiate distinctly between patients and controls.
Exclusion criteria were: studies on specific syndromes, treatment options or mere snorers, case reports, different diagnostic criteria for OSA or AHI threshold, missing control group, fragmentation of patient groups other than gender without declaration of values for the total patient group, patient or control group under 10 participants, insufficiently reported data for statistical analysis.Eligible studies were identified independently by two authors.

| Data extraction
Eligible full texts were screened for evaluated cephalometric measurements and comparable other values.Because of the lack in consistency in the definition of cephalometric measurements, each publication had to be screened for comparability of measured angles and distances.Tables, illustrations, figure captions and full texts were assessed to ensure the comparability of measurements among the included publications, even if they were labelled totally differently according to each country's common style.The different terms used for angles and distances are shown in Table 1.Data extraction was conducted independently by two of the authors.

| Statistical analysis
The statistical analysis was performed with ReviewManager 5.3 (Cochrane Collaboration, London, UK) as random effect models.
Values were included in the meta-analysis if they were comparably reported as mean (SD) in at least four studies.
The risk of bias of the included studies was assessed by two of the authors, in consensus with all authors, using the Quality In Prognosis Studies (QUIPS) tool for the assessment of bias in studies of prognostic factors (Hayden et al., 2006).Heterogeneity was graded according to Higgings and Thompson (Higgins & Thompson, 2002).
The systematic literature research identified 3016 publications meeting the search strategy.A total of 418 publications were excluded according to the title (14 duplicates, 115 case reports, and 289 studies evaluated patients with syndromes), another 2476 after abstract review.The remaining 122 publications were obtained as full text and screened for inclusion and exclusion criteria.In all, 19 publications with 1410 participants met the inclusion criteria and were included in the systematic review and meta-analysis (Figure 1).A total of 16 cephalometric measurements (Figure 2) were sufficiently described and reported to be statistically analysed in the meta-analysis and complemented by age and body mass index (BMI).The included studies and extracted measurements are shown in Table 2.

| Age
Patient age is a well-known risk factor for OSA (Kapur et al., 2017).Eight publications had age-matched controls, two of the included publications (Gungor et al., 2013;Miyao et al., 2000) showed no age difference without explicitly mentioning age-matched controls.The data extracted from 10 included studies and 763 participants without age-matching confirmed this risk factor of a older age being associated with a higher risk of OSA (Figure 3).As the source of patients and controls is not comparable throughout the studies, these findings are probably biased.

| Body mass index
The BMI is also well-known to increase the risk of OSA (Kapur et al., 2017).Four studies had BMI-matched control groups and nine studies with 602 participants were included in the statistic evaluation.
Patients with OSA had a remarkably higher BMI compared to controls (p < 0.00001, Figure 4).To eliminate the impact of patient age on BMI assessment, studies with age-matched or similar-aged controls were separately examined and also showed a higher BMI in the patient groups (p ≤ 0.01, Figure 5).

| Sella-Nasion-A-point angle (SNA)
The sagittal position of the maxilla in relation to the anterior cranial base is commonly measured using the SNA angle according to Graber T A B L E 1 Cephalometric measurements: favoured and alternative terms for angles and distances obtained by lateral cephalometric analysis.(1952) and Steiner (1953).An increased or decreased angle characterises a prognathic or retrognathic maxilla, respectively.In all, 15 studies with 1180 participants were included in the meta-analysis, showing no relevant difference in SNA angles in patients compared to controls (Figure 6).

| Sella-Nasion-B-point angle (SNB)
The sagittal position of the mandible in relation to the anterior cranial base is commonly measured using the SNB angle as described by Brodie (1941).An increased or decreased angle characterises a prognathic or retrognathic mandible respectively.In all, 15 studies with 1181 participants were included in the meta-analysis.Patients showed a slightly decreased SNB angle compared to controls, but without relevance (Figure 7).

| The A-point-Nasion-B-point (ANB) angle
The ANB angle characterises the sagittal relation of maxilla and mandible, described as the difference between SNA and SNB angle by Graber (1952) and Steiner (1953).An increased or decreased angle characterises a skeletal Class II or III, respectively.In all, 12 studies with 894 participants were included in the meta-analysis.Patients with OSA had a higher ANB angle compared to controls (p = 0.02), indicating a distal relation of the mandible relative to the maxilla (Figure 8).

| Sella-Nasion distance (SN)
The anterior cranial base is commonly measured as the distance between the nasofrontal suture (point N) and the middle of hypophyseal fossa (point S) as described by Ricketts (Ricketts, 1957).
Four studies with 273 participants were included in the metaanalysis.Three of the studies examined male adults, one examined children (male and female).Besides this one study with only 30 young patients all studies examined only adult male patients that showed a relevantly shorter anterior cranial base compared to the controls (Figure 9).

| Nasion-Sella-Basion (NSBa) angle
The NSBa angle as described by Brodie (1941)  F I G U R E 3 Impact of age on obstructive sleep apnea (OSA): patients with OSA were relevantly older than controls.
cranial base with a more prognathic face type or a more anteriorly positioned temporomandibular complex.

| Upper anterior face height
The Nasion-anterior nasal spine (N-ANS) distance is one option to describe the upper anterior face height according to Johnson (Johnson, 1950) and was sufficiently measured in six studies with 318 participants.Patients with OSA had a longer upper anterior face height compared to controls ( p = 0.004) with a remarkably low heterogeneity (8%) (Figure 11).Out of three included studies on children (Deng & Gao, 2012;Vieira et al., 2011Vieira et al., , 2014)), two contributed to this effect (Deng & Gao, 2012;Vieira et al., 2011).

| Lower anterior face height
The lower anterior face height can be measured as the distance between ANS and Menton (Me) or Gnathion (Gn) according to Johnson (Johnson, 1950).Seven studies with 342 participants were included in the meta-analysis showing no relevant difference between patients with OSA and controls (Figure 12).

| Total anterior face height
The total anterior face height is commonly measured as the distance between nasion (N) and Me or Gn according to Björk (1947) and Johnson (1950).Five studies with 267 participants were included in the meta-analysis showing no relevant difference between patients with OSA and controls (Figure 13).

| Posterior face height
The distance between Sella (S) and Gonion (Go) is used to describe the posterior face height according to Coben (Coben, 1955).Five studies with 267 participants were included in the meta-analysis.Even if the only study on male patients showed a remarkably shorter posterior face height in patients with OSA, the mean difference over all five studies was almost zero (Figure 14).

| Mandibular line to Nasion-Sella line (ML-NSL) angle
The ML-NSL angle as described by Brodie (Brodie, 1941) describes the rotation of the mandible relative to the anterior cranial base.Eight F I G U R E 5 Impact of body mass index (BMI) on obstructive sleep apnea (OSA) independent of patient age: patients had a higher BMI compared to same-age controls.
studies with 618 participants were included in the meta-analysis.
Patients with OSA showed a relevant posterior rotation of the mandible compared to the controls with a remarkably low heterogeneity (9%) ( p ≤ 0.001, Figure 15).

| The Gn-Go-Ar (Articulare) angle
The mandibular jaw angle or Go angle according to Tweed (1946) and Johnson (1950) is measured between the tangents to the mandible originating in Ar and in Gn.The intersection of these tangents, Go, serves as the vertex of the angle.Four studies with 259 participants were included in the meta-analysis and showed a remarkably more obtuse angle in OSA patients compared to controls with an exceptionally low heterogeneity (0%) (p ≤ 0.01, Figure 16).

| Hyoid bone position
The position of the hyoid bone can be evaluated by the distance to the mandibular plane (MP) according to Riley et al. (1983).In all, 13 studies with 935 participants reported this value and were therefore included in the meta-analysis.All but two of those studies showed a higher distance leading to an overall highly remarkable effect (p ≤ 0.00001, Figure 17).

| Uvula length (UL)
The length of the soft palate or the UL can be measured as the distance between the posterior nasal spine (PNS) and the uvula tip as described by Riley et al. (1983).In all, 13 studies with 1055 participants reported this value and were included in the meta-analysis.

| Uvula thickness (UT)
The thickness of the uvula or the soft palate (UT) is usually measured as the longest distance rectangular to the uvula axis or to F I G U R E 7 Impact of Sella-Nasion-B-point (SNB) angle on obstructive sleep apnea (OSA): there was no relevant difference between patients with OSA and controls.
the UL line as described by Tangugsorn et al. (1995b).This distance was comparably reported in eight studies with 589 participants.Similar to the UL, the thickness of the soft palate was relevantly thicker in patients with OSA compared to controls ( p < 0.00001, Figure 19).

| Posterior airway space (PAS)
Due to the huge variability in measuring the PAS, only two measurements were comparably described and reported in at least four studies respectively.On the B-Go level as described by Riley et al. (1983) F I G U R E 8 Impact of A-point-Nasion-B-point (ANB) angle on obstructive sleep apnea (OSA): patients with OSA had a higher ANB angle indicating a skeletal Class II.

| Risk of bias evaluation and heterogeneity
The risk of bias was evaluated using the QUIPS tool (Hayden et al., 2006).Study participation was graded as a high risk of bias because the source of the population and especially the control groups was consistently described insufficiently.Study attrition was not possible to grade because almost all included studies declared the presence of risk factor (lateral cephalogram) and outcome measurement (OSA diagnosis) as inclusion criteria for the participation in the mostly retrospective studies and did not report drop-outs.Prognostic factor measurement was graded as a high risk of bias in two studies because one (Miyao et al., 2000) adopted the control group from another study without clarifying if the measurements were done by the same operator.The second study (Verin et al., 2002) did not F I G U R E 1 9 Impact of uvula thickness on obstructive sleep apnea (OSA): patients with OSA had a thicker uvula compared to controls.
confounding was graded very differently because some studies had age-, gender-or BMI-matched control groups to avoid the effect of these well-known confounding factors.Other studies recruited the control groups from the institutes' students or staff, that usually provided younger and less obese participants compared to the patient groups.Statistical analysis and reporting were sufficiently described in all included studies, because these were essential inclusion criteria for the present meta-analysis (Supplement 1).
The heterogeneity between the studies was fairly high for all but three measurements (N-ANS, ML-NSL and Gn-Go-Ar), caused by the high distribution based on the differences between the studies rather than on coincidence.
F I G U R E 2 0 Impact of posterior airway space on obstructive sleep apnea (OSA): there was almost no difference detectable concerning the sagittal dimension of the posterior airway space at B-point-Gonion (B-Go) level.
F I G U R E 2 1 Impact of posterior airway space on obstructive sleep apnea (OSA): there was almost no difference detectable concerning the sagittal dimension of the posterior airway space behind the tongue base.
The major challenge while conducting this systematic review and meta-analysis was caused by the inconsistency of labelling cephalometric landmarks, angles, and measurements.The studies that were assessed for inclusion as full texts had up to nine different terms/ abbreviations for the same cephalometric measurement (Table 1) and many studies had to be excluded due to insufficient definition of utilised measurements.This inconsistency significantly impaired the scientific possibilities because important literature can be missed due to not matching terms.Maybe this is one of the reasons why previous systematic reviews on this topic (Agha & Johal, 2017;Armalaite & Lopatiene, 2016;Fagundes et al., 2022;Flores-Mir et al., 2013;Katyal, Pamula, et al., 2013;Miles et al., 1996;Neelapu et al., 2017) did not have a comparable number of publications included in the metaanalysis for the assessment of relevant items, even if the inclusion and exclusion criteria were less strict in some cases compared to the present systematic review.
The approach to measure the PAS is so diverse, that most of the different approaches could not be statistically evaluated in this metaanalysis because there were not at least four includable studies using the same approach.Only two (B-Go level and behind the tongue base) of nine (nasal plane, occlusal plane, ML level, behind the soft palate or the tongue parallel to Frankfort horizontal plane or to B-Go line, respectively) considered measurement approaches were sufficiently comparable.

| Identified risk factors
The results of this meta-analysis reveal relevant hints for several craniofacial measurements being a risk factor for of a side-effect of OSA: The highest relevance was seen for the items UL, UT and the distance between the hyoid bone and the mandibular plane (ML-hyoid bone).
Even if only BMI-matched groups (Akpinar et al., 2011;Pirilä-Parkkinen et al., 2010;Vieira et al., 2011Vieira et al., , 2014) )   is not relevantly increased in children with OSA, while it is seen as a relevant predictive factor in adults, maybe because it is rather a long-term effect of OSA than a causal risk factor.As all primary studies are of a cross-sectional and not a longitudinal design, these aspects need to be evaluated in prospective longitudinal studies in the future.
In summary it can be stated that a thicker and longer soft palate as well as an enlarged distance of the hyoid bone to the mandi- Orthodontic treatment, especially with functional orthopaedic appliances, offer the possibility of growth modification in growing children.The identified items should be kept in mind while treating children in order to avoid negative growth effects and aim for a positive growth modification.

| CONCLUSION
Cephalometric measurements are highly variable world-wide.To facilitate research in this field, a standardisation should be intended in the future, especially concerning the measurements of the PAS, that was almost not comparably described to be statistically evaluated.
Craniofacial anomalies can be a risk factor or an indicator for OSA.The identified most relevant measurements (UL and UT and hyoid bone position, as well as the rotation of the mandible) should be considered in every cephalometric evaluation in order to identify patients with a possibly higher risk of OSA.There findings can improve the screening methods, while the diagnosis has to be determined via polygraphy or PSG.
Many children consult an orthodontist during their growth and may profit by the identification of risk factors and possible growth modification using functional orthopaedic appliances.Lateral cephalograms are commonly used to evaluate the sagittal and vertical orientation of maxilla and mandible, as well as dental arches in relation to the midface and scull.Several angles, distances and proportions are measured, compared to ideal values or individualised norms and subsequently lead to individualised orthodontic treatment plans.They additionally provide information about the two-dimensional width of the upper airway, the thickness of soft tissues and the position of the hyoid bone relative to the spine or the mandible.Lateral cephalometry was developed around 1930 by different working groups simultaneously in the United States and in Germany respectively

F I G U R E 4
Impact of body mass index (BMI) on obstructive sleep apnea (OSA): patients with OSA had a remarkably higher BMI compared to controls.F I G U R E 6 Impact of Sella-Nasion-A-point (SNA) angle on obstructive sleep apnea (OSA): there was no relevant difference between patients with OSA and controls.

F
I G U R E 9 Impact of point Sella-Nasion distance (SN) distance on obstructive sleep apnea (OSA): adult male patients with OSA had a shorter anterior cranial base.F I G U R E 1 0 Impact of Nasion-Sella-Basion (NSBa) angle on obstructive sleep apnea (OSA): patients with OSA showed a sharper NSBa angle compared to controls.F I G U R E 1 1 Impact of Nasion-anterior nasal spine (N-ANS) distance on obstructive sleep apnea (OSA): patients with OSA had a longer upper anterior face height than control groups.288 participants (Figure 20) and behind the tongue base as described by Deberry-Borowiecki et al. (1988) 501 participants (Figure 21) were included in the meta-analysis.Both measurements showed slightly narrower PAS in patients but without relevance.
1 2 Impact of anterior nasal spine-Menton (ANS-Me) distance on obstructive sleep apnea (OSA): there was no relevant difference between patients and controls concerning the lower anterior face height.F I G U R E 1 3 Impact of total anterior face height on obstructive sleep apnea (OSA): there was no relevant difference between patients and controls.provide a figure to clarify the cephalometric measurements and the full text only comprised the basic information to allow the inclusion for this systematic review.Outcome measurement was graded with a medium risk of bias because not all studies conducted a polygraphy or PSG on control groups to exclude the possibility of including patients with minor OSA but without subjective health problems.The study F I G U R E 1 4 Impact of Sella-Gonion (S-Go) distance on obstructive sleep apnea (OSA): there was almost no difference in posterior face height in patients with OSA compared with controls.F I G U R E 1 5 Impact of the mandibular line to Nasion-Sella line (ML-NSL) angle on obstructive sleep apnea (OSA): patients with OSA showed a higher posterior rotation of the mandible relative to the anterior cranial base compared to controls.F I G U R E 1 6 Impact of Gnathion-Gonion-Articulare (Gn-Go-Ar) angle on obstructive sleep apnea (OSA): patients with OSA had a more obtuse jaw angle indicating a vertical growth pattern.F I G U R E 1 7 Impact of mandibular plane-hyoid bone (MP-H) distance on obstructive sleep apnea (OSA): patients with OSA showed a greater distance between the hyoid bone and the mandibular plane.F I G U R E 1 8 Impact of uvula length on obstructive sleep apnea (OSA): patients with OSA had a longer uvula compared to controls.
to a relevant risk of bias concerning these two parameters and possibly also other confounding factors.The sagittal relation of the mandible to the maxilla and the vertical rotation of the mandible, described by ANB angle, ML-NSL angle and Gn-Go-Ar angle, seem to be a risk factor for OSA as well as the upper anterior face height, the anterior cranial base length SN and inclination NSBa.The mean differences of these parameters are relatively low compared to the total values and commonly accepted normal ranges, which reduces the clinical relevance.Due to the simultaneous assessment of prognostic factors (lateral cephalogram) and outcome measurements (OSA diagnosis) in many of the studies, a differentiation concerning causality is not clear.There is no reliable information about the onset of OSA, thus the supposed prognostic factor could also be a consequence of the disease.Nevertheless, the observation of these craniofacial modifications may lead to an earlier diagnosis of OSA and maybe avoid its implications.Only four of the 19 studies included in this meta-analysis examined children.There is a variety of studies and even meta-analyses on facial features of snoring children but only a few provide data of children with OSA diagnosed by polygraphy or PSG and lateral cephalograms.Interestingly, craniofacial features like a skeletal Class II and a posterior mandibular rotation-represented by increased ANB and ML-NSL angles-are present in children with OSA already, which might lead to the conclusion, that these are rather aetiological factors than consequences.In contrast, the length of the soft palate ble are relevant cephalometric indicators for OSA, as well as a posterior rotation of the mandible and an enlarged upper anterior facial height.A skeletal Class II and a shorter and a more posteriorly inclined anterior cranial base seem to also indicate a higher risk of OSA.Those craniofacial features lead to a convex profile, that might as well be detectable for paediatricians without X-ray imaging and encourage the referral to a sleep laboratory if in doubt.