"CraniOfaCial pHEnOTyping Of OSa Craniofacial Phenotyping in Obstructive Sleep Apnea – A Novel QuantitativePhotographic Approach 1,2 1,2 2 1,2 Richard W. W. Lee, MD ; Andrew S. L. Chan, MD ; Ronald R. Grunstein, MD, PhD ; Peter A. Cistulli, MD, PhD 1 2 Centre for Sleep Health and Research, Department of Respiratory Medicine, Royal North Shore Hospital, NSW, Australia; Woolcock Institute ofMedical Research, University of Sydney, NSW, Australia Study Objectives: To compare the craniofacial morphological pheno- and neck. After 1-for-1 subgroup matching for BMI and sex (51 subjectstype of subjects with and without obstructive sleep apnea (OSA) using in each group), mandibular length 1 (6.21 ± 0.08 [mean ± SEM] vs 6.58a quantitative photographic analysis technique. ± 0.08 cm, P = 0.006), mandibular-nasion angle 1 (35.0 ± 0.48 vs 36.7Design: Case-control study; subgroup matched for body mass index ± 0.37 degrees, P = 0.006) and anterior neck space area (10.2 ± 0.53 vs2 (BMI) and sex. 12.2 ± 0.52 cm , P = 0.01) remained smaller in the OSA group. MandibularSetting: Sleep investigation unit in a university teaching hospital. width-length angle (88.0 ± 0.75 vs 85.3 ± 0.54 degrees, P = 0.005) andpatients: 114 subjects (93% Caucasian) with OSA (apnea-hypopnea face width-midface depth angle (72.3 ± 0.44 vs 70.7 ± 0.39 degrees, P =index [AHI] = 10/h) and 66 controls (AHI < 10/h). 0.01) remained larger in the OSA group, whereas mandibular triangular2 interventions: Standardized frontal-profle craniofacial photographic area (39.2 ± 0.63 vs 41.7 ± 0.74 cm , P = 0.01) was smaller. imaging performed prior to polysomnography. Photographs were ana- Conclusions: Craniofacial phenotypic differences in OSA in Caucasianlyzed for the computation of linear, angular, area and polyhedral volume subjects can be demonstrated using a photographic analysis technique. measurements representing dimensions and relationships of the vari- Keywords: obstructive sleep apnea; craniofacial abnormalities; photo- ous craniofacial regions. grammetry; phenotype Measurements and results: Photographic craniofacial phenotypic differ- Citation: Lee RWW; Chan ASL; Grunstein RR; Cistulli PA. Craniofacialences were demonstrated between OSA and control subjects, including a phenotyping in obstructive sleep apnea – a novel quantitative photo- range of measurements of the face, mandible, maxilla, eyes, nose, head graphic approach. SLEEP 2009;32(1):37-45. OBSTRUCTIVE SLEEP APNEA (OSA) IS CHARACTER- While the available techniques (cephalometry, computed to- IZED BY REPETITIVE CLOSURE OF THE UPPER AIRWAY mography [CT] and magnetic resonance imaging [MRI]) forDURING SLEEP. ITS OCCURRENCE IS THE RESULT OF craniofacial assessment allow detailed examination of bony andanatomic and functional abnormalities of the upper airway. While soft tissue structures, they are generally limited to research ap- neuromuscular and respiratory control mechanisms play impor- plications due to their expense, time-consuming analyses and/ tant roles in the maintenance of airway patency, an abnormal or radiation exposure. Craniofacial anthropometry and photo- 1,2 anatomic substrate is a key factor in the development of OSA. grammetry are alternative craniofacial assessment techniquesCraniofacial abnormalities, enlargement of upper airway soft tis- that have the advantages of being noninvasive and readily ac- sue structures, central obesity, and an excess of regional adipose cessible. Furthermore, they allow quantifcation of the surface3-5 tissue are known anatomic risk factors for OSA. Although obe- morphology which is generally not achievable with the othersity is generally considered the major attributing risk factor for imaging modalities. Craniofacial photogrammetry, involving6 OSA, craniofacial morphology is increasingly recognized as an measurements from photographs, has been applied in the as- 17,18 important interacting factor in OSA pathogenesis. sessment of subjects with craniofacial anomalies. Applica- It is well recognized from studies using imaging techniques tion of photogrammetry to examine subjects with OSA may re- that craniofacial abnormalities are common in patients with veal new insights in the craniofacial morphological phenotype7-9 OSA. Mandibular retrusion, maxillary defciency, inferior of this condition. displacement of the hyoid bone and cranial base abnormalities The aims of this study were to compare the craniofacial mor- 7,8,10,11 are amongst the most commonly reported fndings. These phological phenotype of subjects with and without OSA usingabnormalities can result in a compromised airway space and an a quantitative photographic analysis technique. Secondly, we12,13 increase in upper airway collapsibility, thereby predisposing aimed to determine whether these differences were presentto OSA. It is thought that the interaction between craniofacial between subjects with similar degrees of obesity. Finally, wemorphology and obesity determines the likelihood and severity aimed to examine the relationships of these craniofacial photo- 14-16 of OSA in the majority of patients. graphic measurements to obesity and OSA severity. METHODS Submitted for publication June, 2008 Submitted in fnal revised form August, 2008 Subjects accepted for publication august, 2008 Address correspondence to: Peter Cistulli, MD, PhD, Centre for SleepSubjects referred for polysomnography to a university teach- Health and Research, Department of Respiratory Medicine, Level 8, Maining hospital for the initial investigation of OSA were recruitedBlock, Pacifc Highway, St Leonards, NSW 2065, Australia; Tel: +61 2consecutively. Exclusion criteria included those with a known9926 8674; Fax: +61 2 9906 6391; E-mail: [email protected] SLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 37along the subject alignment plane while ensuring both earswere seen equally from the front. For the profle photograph,the subject was instructed to turn 90 degrees to the left after thefrontal photograph was taken. This was aided by a laser pointerhead-clip and calibrated markings on the side wall to ensurethe profle views were perpendicular to the frontal views. Thesubject’s mid-sagittal plane was aligned to the subject align- ment plane. Using image analysis software (Image J v1.36, NIH, Bethes- da, MD), the photographs were examined for landmark digitiza- tion. Craniofacial landmarks of interest were captured as pixelcoordinates (x, y) of the image which were then transferred toa custom-programmed spreadsheet for the computation of lin- ear, angular, area, and polyhedral volume measurements. Pixelmeasurements were converted to metric dimensions (52 pixels/ Figure 1—Photographic Landmarks – Profle and Frontal View.cm). In this study, a total of 71 craniofacial measurements wereLandmarks pre-identifed on subject (marked with a white tape) :computed using this photogrammetry technique. These mea- sup – infraorbital rim; me – mentum; ty – thyroid; cr – cricoid;surements represented the dimensions and relationships of theste – sternal notch; gol – gonion (L); gor – gonion (R). Land- marks digitized on photographs: t – tragion; ex – exocanthion; various craniofacial regions including the face, mandible, max- sup – infraorbital rim; g – glabella; n – nasion; sn – subnasion; stoilla, eyes, nose, head, and neck. Technique validation (landmark– stomion; sl – sublabiale; gn – gnathion; me – mentum; cer – cer- digitization accuracy and test-retest reliability) was performedvical point; ty – thyroid; cr – cricoid; np; neck plane; ste – sternalin a subgroup of subjects. notch; go – gonion; ra – ramus; op – opisthocranion; v – vertex;aneck – anterior neck; pneck – posterior neck; tl – tragion (L); tr –anthropometry tragion (R); gol – gonion (L); gor – gonion (R); eul – euryon (L);eur – euryon (R); exl – exocanthion (L); exr – exocanthion (R);Height was measured by a wall-mounted stadiometer (±0.1enl – endocanthion (L); enr – endocanthion (R); lal – alare (L);cm). Subjects were weighed using an analogue scale (±0.5 kg)ral – alare (R); lneck – neck (L); rneck – neck (R); (L) = left sideon the photograph, (R) = right side on the photograph. with minimal clothing. Neck circumference was measured witha tape measure (±0.5 cm) at the level of the cricothyroid mem- brane. Waist circumference was measured at the level of thehistory of syndromal craniofacial abnormalities (e.g., Down ischial tuberosities with the subject in the standing position.syndrome), previous craniofacial surgery, and excessive facial Body mass index (BMI) was calculated with the formula of2 hair which signifcantly obscured facial landmarks. Subjects of weight (kg) divided by height squared (m ). all ethnicity (self-reported) were included. Anthropometry andthe photographic procedure were performed on all subjects on polysomnography the same day as the polysomnography. All data collection andphotographic analyses were carried out by a single investiga- Diagnostic polysomnography (PSG) was performed in ac- 19,20 tor (RL) who was blinded to the result of the polysomnogra- cordance with previous studies and recommendations. Sleep21 phy. Ethics approval was obtained from the institutional ethics staging was determined using standardized defnitions. Apneacommittee, and written informed consent was obtained from all was defned as complete airfow cessation = 10 seconds with ox - subjects. ygen desaturation of at least 3% and/or associated with arousal.Hypopnea was defned as a reduction in amplitude of airfow orStandardized photographic Technique and Craniofacial thoracoabdominal wall movement > 50% of the baseline mea- photogrammetry surement > 10 sec with an accompanying oxygen desaturationof at least 3%, and/or associated with arousals. Apnea-hypopneaFrontal and profle digital photographs of the head and neck index (AHI) was calculated as the total number of apneas andwere obtained with a standardized setup. A single-lens refex hypopneas per hour of sleep. Polysomnography scoring wasdigital camera (D70 with 18-70 mm lens and external fash unit performed by experienced accredited sleep technologists. TheSB-29s; Nikon Corp., Japan) was mounted on a tripod at a dis- OSA cases were defned by an AHI = 10 events per hour. Thetance of 160 cm from the subject alignment plane. Standard- controls were defned by an AHI < 10 events per hour. ized camera settings (focal length 70 mm, aperture 7.1, shutterth speed 1/100 , ISO 400) were used to ensure consistency of the Data and Statistical analysis JPEG images (resolution 3008 by 2000 pixels). Subjects werephotographed standing upright while assuming the natural head Statistical analysis was performed with SPSS (v13.0 for2 position. Prior to the photographs, certain bony and cartilagi- Windows, SPSS Inc., Chicago, IL, USA). The ? test was usednous landmarks were pre-identifed on the subjects by palpation for comparing categorical variables (gender and ethnicity).and marked with a white tape (Figure 1). Standardized methods Fisher exact test was used if any of the cells in a categoricalwere used to align subjects for the photographs. For the frontal table have expected count less than 5. Student’s t-test was usedphotograph, the subject’s facial landmark nasion was aligned for comparing normally distributed continuous variables (cran- SLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 38Table 1—Subject Characteristics Control OSA P \t \t (AHI\t<\t10)\t (AHI\t=\t10)\t n = 66 n = 114Males (%) 46 (69.7%) 91 (79.8%) NS Age (years) 49.3 ± 14.9 55.8 ± 13.5 0.003 Ethnicity (% Caucasians) 100% 93% 0.02 Anthropometry2BMI (kg/m ) 27.1 ± 4.8 30.6 ± 4.9 < 0.001Height (cm) 173.7 ± 9.6 172.7 ± 9.3 NSWeight (kg) 82.6 ± 19.1 91.2 ± 15.6 0.002Neck circumference (cm) 39.5 ± 4.5 42.4 ± 4.1 < 0.001Waist circumference (cm) 97.5 ± 14.2 109.4 ± 12.7 < 0.001 Epworth sleepiness scale 8.7 ± 0.61 8.9 ± 0.47 NS Polysomnography Total AHI 4.4 ± 3.1 33.4 ± 20.8 -AHI – NREM sleep 3.5 ± 2.8 31.6 ± 22.6 -AHI – REM sleep 8.3 ± 8.7 37.3 ± 22.8 -Minimum SaO (%) 90.4 ± 4.8 79.5 ± 10.2 - 2Arousal index 22.6 ± 11.5 40.6 ± 17.3 -Sleep effciency (%) 79.1 ± 14.6 78.6 ± 16.3 - Mean ± SD; NS = nonsignifcant; BMI = body mass index; AHI = apnea-hypopnea index iofacial photogrammetry data). One-for-one matching for BMI ground, with the exception of 5 Chinese and 3 Pacifc Islander2 (±0.5 kg/m ) and sex was performed in a subgroup of OSA and subjects. control subjects. Analysis of the relationship between craniofacial photo- Craniofacial photogrammetry grammetry and OSA severity, and obesity (BMI, neck circum- ference, waist circumference) were limited to those measure- Results of the 71 craniofacial photographic measurementsments which had a P value = 0.10 in the primary case control comparing OSA subjects and controls are presented in Table 2.analysis (to reduce the number of statistical comparisons). The These results are summarized below according to the variousAHI (with the addition of one) was logarithmic-transformed to craniofacial regions. obtain a normal distribution. Pearson correlations were used toexamine the association between continuous variables. A par- face tial correlation procedure was used to examine the linear rela- tionship between OSA severity and craniofacial photographic There were no signifcant differences in the vertical dimen - measurements while controlling for the effect of BMI. sions of the face (total, upper, or lower face heights). The midA P value = 0.01 was considered statistically signifcant. All and lower face widths were signifcantly greater in subjectsP values < 0.10 were shown in the results, those that were great- with OSA (face width, mandibular width). In addition, theer than or equal to 0.10 were shown as nonsignifcant (NS). All mid and lower face appeared to be wider and fatter ( mandibu- data are expressed as mean ± SEM unless otherwise specifed. lar width-length angle, face width-midface depth angle, facewidth-lower face depth angle). The axial triangular or polygo- rESUl TS nal areas of the face at the level of the cranial base or maxillawere larger in the OSA subjects (cranial base triangle areaSubjects Characteristics (ax); cranial base area 1 (ax) and 2 (ax); maxillary trianglearea (ax)). Similarly, the volume of the midface region wasA total of 180 subjects were recruited and included in the also larger in the OSA subjects (middle cranial fossa volume,analysis; this included 114 subjects in the OSA group (AHI = maxillary volume). On the contrary, the areas and volume of10) and 66 control subjects (AHI < 10). Three subjects were ex- the face in the mandibular region were similar between thecluded from the analysis (2 subjects did not complete the PSG; 2 groups (mandibular triangle area and mandibular triangle1 subject was found to have central sleep apnea). Twelve sub- area (ax); mandibular volume). The facial axis angle wasjects (7 males, 5 females) declined study participation. Char- similar between the 2 groups. acteristics of the subjects in the control and OSA groups aresummarized in Table 1. There was no signifcant difference in Mandible and Maxilla the proportion of males in the comparison groups. The meanage was higher in the OSA subjects (P = 0.003), as were various The length of the mandible was shorter in the subjects withanthropometric measures of obesity (ranged from P = 0.002 to OSA (mandibular length 1 and 2). Correspondingly, a numberP < 0.001). The majority of subjects were of Caucasian back- of angular measurements also suggest the mandible was shorterSLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 39Table 2—Craniofacial Photogrammetry – Primary Analysis Craniofacial Control OSA P \t \t Landmarks\t (AHI\t<\t10)\t (AHI\t=\t10)\tn = 66 n = 114Face*Upper face depth t-n 9.93 ± 0.08 10.0 ± 0.06 NSMid face depth 1 t-sn 10.5 ± 0.08 10.5 ± 0.06 NSMid face depth 2 t-sl 11.4 ± 0.09 11.5 ± 0.06 NSLower face depth 1 t-gn 12.9 ± 0.10 13.0 ± 0.07 NSLower face depth 2 t-me 12.4 ± 0.11 12.5 ± 0.07 NSTotal face height n-gn 11.9 ± 0.09 12.1 ± 0.07 0.03Upper face height n-sto 7.72 ± 0.06 7.89 ± 0.05 0.03Lower face height 1 sn-gn 6.52 ± 0.08 6.65 ± 0.06 NSLateral face height ex-go 10.3 ± 0.10 10.7 ± 0.07 0.005Face width tl-tr 15.0 ± 0.09 15.7 ± 0.08 < 0.001Facial axis angle n-t and go-gn 36.2 ± 0.84 35.2 ± 0.64 NSMandibular width-length angle gor-me-gol 84.4 ± 0.56 89.4 ± 0.58 < 0.001Face width-midface depth angle tr-sn-tl 70.5 ± 0.37 72.7 ± 0.32 < 0.001Face width-lower face depth angle tr-me-tl 61.9 ± 0.31 63.8 ± 0.28 < 0.001Maxillary triangle area t-sn-me 39.0 ± 0.66 39.9 ± 0.42 NSMandibular triangle area t-go-me 15.3 ± 0.57 15.3 ± 0.42 NSMaxillary-mandibular box area t-sn-me-go 54.4 ± 0.99 55.2 ± 0.66 NSCranial base triangle area (ax) tl-n-tr 74.7 ± 0.87 78.6 ± 0.71 0.001Cranial base area 1 (ax) tl-exl-exr-tr 96.2 ± 1.04 102 ± 0.91 < 0.001Cranial base area 2 (ax) tl-exl-n-exr-tr 104 ± 1.12 110 ± 0.97 < 0.001Maxillary triangle area (ax) tl-sn-tr 78.9 ± 0.90 82.5 ± 0.77 0.003Middle cranial fossa volume tl-tr-n-sn 134 ± 2.21 143 ± 1.81 0.002Maxillary volume tl-tr-sn-me 196 ± 3.97 209 ± 2.84 0.007Mandibular volume tl-tr-gol-gor-me 141 ± 5.88 148 ± 4.49 NSMaxillary-mandibular volume 1 tl-tr-gol-gor-sn-me 337 ± 8.47 357 ± 6.31 0.06 Mandible and Maxilla*Anterior mandibular height sto-gn 4.19 ± 0.05 4.27 ± 0.04 NSMandibular length 1 me-go 6.54 ± 0.07 6.19 ± 0.06 < 0.001Mandibular length 2 gn-go 7.57 ± 0.08 7.29 ± 0.06 0.008Posterior mandibular height t-go 6.86 ± 0.10 7.31 ± 0.07 < 0.001Mandible width gol-gor 12.3 ± 0.12 13.0 ± 0.11 < 0.001Maxillary depth angle t-n-sn 80.1 ± 0.41 79.2 ± 0.44 NSMandibular depth angle 1 t-n-sl 72.3 ± 0.43 71.9 ± 0.41 NSMaxillary-mandibular relationship angle 1 sn-n-sl 7.87 ± 0.35 7.36 ± 0.28 NSMandibular-nasion angle 1 go-n-gn 37.0 ± 0.37 34.8 ± 0.32 < 0.001Mandibular-nasion angle 2 go-n-me 30.6 ± 0.34 28.2 ± 0.30 < 0.001Mandibular-subnasion angle 1 go-sn-gn 55.9 ± 0.58 53.2 ± 0.52 0.001Mandibular-subnasion angle 2 go-sn-me 45.8 ± 0.48 42.7 ± 0.45 < 0.001Mandibular plane angle 1 go-me-TH 29.1 ± 1.05 27.7 ± 0.94 NSMandibular triangle area (ax) gol-me-gor 40.5 ± 0.71 40.3 ± 0.52 NS Eyes and NoseNose height n-sn 5.47 ± 0.05 5.58 ± 0.04 0.09Eye width exl-enl 2.73 ± 0.03 2.68 ± 0.03 NSIntercanthal width enl-enr 3.27 ± 0.05 3.50 ± 0.04 < 0.001Biocular width exl-exr 8.73 ± 0.05 8.87 ± 0.05 0.06Nose width lal-ral 3.72 ± 0.05 3.97 ± 0.04 < 0.001 HeadTotal craniofacial height v-gn 24.8 ± 0.18 24.8 ± 0.12 NSMaximum cranial length g-op 20.6 ± 0.17 20.7 ± 0.10 NSMaximum cranial width eul-eur 14.4 ± 0.12 14.7 ± 0.08 0.02Natural Head position angle t-sup-TH 3.68 ± 0.66 4.62 ± 0.61 NSHead base Inclination angle t-n-TH 16.5 ± 0.71 17.6 ± 0.63 NS NeckThyromental distance – horizontal ty-me (TH) 4.09 ± 0.10 4.12 ± 0.08 NSCricomental distance – horizontal cr-me (TH) 4.91 ± 0.11 5.04 ± 0.08 NSSternomental distance ste-me 9.74 ± 0.16 9.42 ± 0.15 NSSternomandibular distance – vertical ste-go (TV) 11.6 ± 0.14 10.8 ± 0.16 < 0.001 (continued on following page) SLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 40Table 2—Craniofacial Photogrammetry – Primary Analysis (continued) Craniofacial Control OSA P \t \t Landmarks\t (AHI\t<\t10)\t (AHI\t=\t10)\tn = 66 n = 114 Sternotragion distance – vertical ste-t (TV) 18.1 ± 0.14 17.6 ± 0.15 0.05Cricomental space distance† cer-cr-me 0.65 ± 0.06 0.29 ± 0.05 < 0.001Neck depth aneck-pneck 12.6 ± 0.21 13.9 ± 0.14 < 0.001Neck width lneck-rneck 12.3 ± 0.16 13.1 ± 0.12 < 0.001Neck perimeter l-r-a-p-neck 39.1 ± 0.55 42.4 ± 0.38 < 0.001Cervicomental angle np-cer-me 154 ± 2.20 167 ± 1.64 < 0.001Cricomental space area† cer-cr-me 2.15 ± 0.22 1.01 ± 0.16 < 0.001Thyromental space area† cer-ty-me 1.02 ± 0.15 0.35 ± 0.10 < 0.001Anterior neck space area† ste-cr-ty-cer-me 12.3 ± 0.47 10.2 ± 0.41 0.001Total anterior neck soft tissue area go-me-cer-cr 18.0 ± 0.43 18.2 ± 0.31 NSPosterior neck soft tissue area cr-go-pneck 39.9 ± 0.81 43.1 ± 0.73 0.006Total neck soft tissue area go-me-cer-cr-pneck 58.0 ± 1.13 61.3 ± 0.93 0.03Neck cross-sectional area l-r-a-p-neck 123 ± 3.38 144 ± 2.64 < 0.001Mandibular cricoid area (ax) gol-gor-cr 47.2 ± 0.72 48.0 ± 0.69 NSTotal anterior neck soft tissue volume 2‡ gol-gor-me-cer-ty-cr 74.0 ± 2.15 78.6 ± 1.59 0.08Total anterior neck space volume 2† (gol-gor-me-cr) – ‡ 9.27 ± 0.82 4.98 ± 0.72 < 0.001Posterior neck soft tissue volume§ pneck-cr-gol-gor 494 ± 13.1 566 ± 12.9 < 0.001Total neck soft tissue volume ‡ + § 567 ± 14.7 645 ± 14.0 < 0.001 Mean ± SEM. NS = nonsignifcant; TH = true horizontal distance between 2 landmarks; TV = true vertical distance between 2 landmarks; ax= axial measurement. *Photographic measurements of the face and maxillomandibular region can overlap; †Measurement can be of a negative2 3 value; Linear measurements are in cm, angles in degrees, areas in cm , volumes in cm . in these subjects (mandibular-nasion angle 1 and 2; mandibular- were all greater in the OSA subjects (neck width; neck depth;subnasion angle 1 and 2). Maxillary defciency was not detected neck perimeter; neck cross-sectional area; posterior and totalin the OSA subjects (mid face depth 1 and 2; maxillary depth neck soft tissue volume). While the anterior neck soft tissue areaangle; maxillary-mandibular relationship angle 1). The inferior and volume were similar (total anterior neck soft tissue area;plane of the mandibular body relative to the horizontal plane was total anterior neck soft tissue volume 2), the area and volumesimilar between the groups (mandibular plane angle 1). of space in front of the neck and below the jaw were signif - cantly smaller in the OSA subjects (anterior neck space area;Eyes and nose total anterior neck space volume 2). Measurements that relateto the cricoid cartilage, thyroid cartilage and soft tissues on theThere was no difference in the horizontal size of the eyes anterior neck region were also different between the compari- between subjects with OSA and controls (eye width). However, son groups (cricomental space distance; cervicomental angle;the distance between the inner corners of the eyes was greater cricomental space area; thyromental space area). (intercanthal width) in subjects with OSA, but not between theouter corners of the eyes (biocular width). There was no differ- Subgroup analysis: Craniofacial photogrammetry in Caucasianence in the nose height, but the nose width was greater in the Men OSA subjects. Analysis was performed in the subgroup of 131 CaucasianHead men (85 OSA vs 46 controls). Differences in the craniofacialphotographic measurements were consistent with those ob- The vertical or anteroposterior lengths of the head were not tained in the primary analysis (data not shown). different between subjects with OSA and controls (total cran- iofacial height; maximum cranial length). Neither was there Subgroup analysis: Craniofacial photogrammetry in BMi andany signifcant difference in the lateral dimension of the head Sex Matched Subjects (maximum cranial width). Head position relative to the hori- zontal plane was similar between the two groups (natural head Subgroup analysis was performed after matching OSA andposition angle; head base inclination angle). control subjects for BMI and sex. This procedure resulted in 51subjects in each comparison group (Table 3). The mean age wasneck higher in the OSA group (P = 0.01). No signifcant differenceswere seen in the neck circumference and waist circumferenceThe vertical length of the neck tended to be shorter in the between the two groups. Compared to the primary analysis ofOSA group (sternomandibular distance – vertical; sternotra- the craniofacial photogrammetry data, a number of measure- gion distance – vertical). Measurements that relate to neck size ments were no longer different between the control and OSASLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 41Table 3—Craniofacial Photogrammetry – Subgroup Analysis 1-for-1 Matched for BMI and Sex Control OSA P \t \t (AHI\t<\t10)\t (AHI\t=\t10)\t n = 51 n = 51Total AHI* 4.7 ± 3.1 29.0 ± 18.4 - Males (%) 40 (78.4%) 40 (78.4%) - Age (years)* 49.5 ± 14.6 56.5 ± 13.0 0.01 Anthropometry2BMI (kg/m )* 28.4 ± 4.62 28.4 ± 4.08 -Neck circumference (cm)* 40.8 ± 4.09 40.9 ± 2.96 NSWaist circumference (cm)* 102 ± 13.0 104 ± 9.40 NS Craniofacial Photogrammetry Mandibular length 1 6.58 ± 0.08 6.21 ± 0.08 0.001Mandibular length 2 7.62 ± 0.08 7.23 ± 0.09 0.002Mandibular-nasion angle 1 36.7 ± 0.37 35.0 ± 0.48 0.006Mandibular-nasion angle 2 30.3 ± 0.34 28.6 ± 0.45 0.004Mandibular width-length angle 85.3 ± 0.54 88.0 ± 0.75 0.005Face width-midface depth angle 70.7 ± 0.39 72.3 ± 0.44 0.01Anterior neck space area† 12.2 ± 0.52 10.2 ± 0.53 0.01Mandibular triangle area (ax) 41.7 ± 0.74 39.2 ± 0.63 0.01 *Mean ± SD; NS = nonsignifcant; Craniofacial photogrammetry results with P = 0.01 are shown; ax = axial measurement; †Measurement can2 be of a negative value; Linear measurements are in cm, angles in degrees, areas in cm groups after matching for BMI and sex (see Tables 2 and 3). relationships remained (e.g., face width [r = 0.36, P < 0.001],However, mandibular length 1 (P = 0.001) and mandibular mandibular width [r = 0.28, P < 0.001]). length 2 (P = 0.002) remained shorter; mandibular-nasion angle1 (P = 0.006), mandibular-nasion angle 2 (P = 0.004) and an- Standardized photographic Technique Validation terior neck space area (P = 0.01) remained smaller, in the OSAgroup. The mandibular width-length angle (P = 0.005) and face Landmark digitization accuracy and test-retest reliabilitywidth-mid face depth angle (P = 0.01) remained larger in the were assessed in 20 subjects who completed the photographicOSA group. In contrast to the primary analysis, the mandibular imaging on two separate occasions with photogrammetry per- triangle area (ax) (P = 0.01) was smaller in the OSA group. formed on separate days for each set of photographs. The over- all mean coeffcient of variation (CV) was 3.45% and intraclassrelationship to Obesity correlation coeffcient (ICC) was 0.96 for all the craniofacialmeasurements. Linear relationships between craniofacial photogrammetryand anthropometric measures of obesity were examined in the DiSCUSSiOn entire cohort of 180 subjects. Other than the photographic mea- surements relating directly to the neck (e.g., neck perimeter, This study demonstrates that craniofacial differences be- neck cross-sectional area, etc.), the face width (r = 0.52, P < tween Caucasian subjects with and without OSA can be identi- 0.001), mandible width (r = 0.58, P < 0.001) and cervicomen- fed with a novel craniofacial photographic analysis technique.tal angle (r = 0.50, P < 0.001) had the strongest relationships These differences include a range of measurements represent- with BMI. Similarly, the face width and mandible width had ing the morphological phenotype of the various craniofacialthe strongest relationships with neck circumference (r regions. Furthermore, these phenotypic differences were dem- [face width] = 0.76, P < 0.001; r = 0.76, P < 0.001) and waist onstrated independent of obesity. This study also suggests that[mandible width] circumference (r = 0.66, P < 0.001; r = 0.70, some of these photographic measurements are related to mea- [face width] [mandible width] P < 0.001). sures of obesity and OSA severity. Photographic differences were demonstrated across all therelationship to OSa Severity craniofacial regions including the face, mandible, maxilla, eyes,nose, head, and neck. Notably, most of these measurements ap- Linear relationships between craniofacial photogrammetry pear to be capturing a combination of regional soft tissues andand OSA severity were also examined in the entire cohort of skeletal anatomy. For example, the face width, mandible width,subjects. The strongest relationships were shown with the neck intercanthal width, and nose width measurements were all largerdepth (r = 0.51, P < 0.001), neck perimeter (r = 0.50, P < 0.001), in the OSA subjects in the primary analysis, whereas they wereneck cross-sectional area (r = 0.49, P < 0.001), face width (Fig- no longer signifcantly different in the BMI and sex matchedure 2a; r = 0.49, P < 0.001), mandible width (Figure 2b; r = subgroup analysis. Furthermore, there were moderate to strong0.45, P < 0.001) and mandibular width-length angle (Figure 2c; linear relationships between some of these measurements andr = 0.45, P < 0.001). After controlling for BMI, these positive anthropometric measures of obesity. These data strongly sup- SLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 42mid and lower face, and more soft tissues or fat deposition onFigure 2a the anterior neck. Collectively, these fndings demonstrate theaverage craniofacial phenotype of subjects with OSA, withoutthe infuence of obesity. In general, these craniofacial phenotypic fndings refectthose identifed previously using standard imaging modalities.Retrusion or shortening of the mandible on cephalometry is7,8 one of the most consistent skeletal fnding in OSA. A widermandibular divergence (i.e., wider and fatter lower face) andsmaller internal area bounded by the mandible have also been9 shown using MRI. Increased total neck size is well establishedr = 0.49as an independent risk factor for OSA, even after controlling forP < 0.0015,22 obesity. Specifcally, distribution of fat within the neck seems23 to localize to the anterolateral aspects. This corresponds to thereduced area of space anterior to the neck found in the photo- graphic analysis. Figure 2b Craniofacial surface morphology assessment in OSA hasmainly been restricted to using anthropometric techniques.Consequently, the range of craniofacial features examined hasbeen very limited (e.g., retrognathia on clinical examination,cranial dimensions with calipers). While retrognathia may bea consistent fnding in OSA subjects using cephalometry, it ap - 4,24 pears not to be detectable on clinical examination. This mightsuggest anthropometry and non-quantitative methods lack sen- sitivity in detecting craniofacial differences in OSA. Surfacemeasurements of the anterior neck, such as the thyromental an- r = 0.45 gle and cricomental space, have been examined and appear toP < 0.00125,26 be different between subjects with and without OSA. Thesefndings are consistent with the photographic differences (e.g.,cervicomental angle, cricomental space area) demonstrated inthe primary analysis. Limited work has been performed examining lateral dimen- Figure 2c sions of craniofacial structures. Cephalometry studies weregenerally limited to profle imaging, therefore only allowingassessment of anteroposterior (AP) dimensions and relation- ships. Similarly, MRI or CT imaging studies did not specifcallyexamine for lateral craniofacial dimensions. Anthropometricsof the craniofacial form in subjects with OSA, however, do sug- gest that Caucasian OSA subjects have brachycephalic (widerlaterally and shorter AP dimension) head form and eurypros- 3 opic (wider laterally and shorter vertically) facial form. Thelatter appears to be consistent with the photographic analysisr = 0.45 which suggests that OSA subjects have a wider and fatter mid - P < 0.001face. Together with a wider and fatter face at the level of themandible, we speculate these phenotypic fndings could refectthe combined effects of excess facial fat/soft tissues and maxil- lary/mandibular defciency. Figure 2—Relationships between OSA Severity (Log [AHI + 1])Skeletal abnormalities, such as mandibular retrusion and max- and Craniofacial Photographic Measurements illary defciency, can result in a compromised airway space and12,13 an increase in upper airway collapsibility, thereby predispos- ing to OSA. While obesity might be strongly linked to OSA, thisport the notion that some craniofacial photographic measure- relationship is mediated through regional body fat distribution.ments are capturing facial soft tissues or adiposity, which are in Increased abdominal or visceral fat may cause upper airwayturn closely linked to general and regional obesity. compromise by reducing chest wall compliance, reducing lung6 The BMI and sex-matched subgroup analysis demonstrated volumes and hormonal factors. Increased neck fat might causea number of craniofacial differences between subjects with airway collapse by the direct mass loading effect in the supine27 and without OSA, independent of the effect of general obesity. position. However, facial fat deposition and OSA have not pre- Typically, subjects with OSA have a shorter and retruded jaw, viously been examined. On the basis of the photographic pheno- smaller enclosed area within the mandible, wider and fatter type fndings, we speculate that facial adiposity may be an im - SLEEP, Vol. 32, No. 1, 2009 Craniofacial Phenotyping in OSA—Lee et al 43"