INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Obstructive sleep apnea (OSA), a condition caused by pharyngeal obstruction during sleep, is manifest as arterial oxygen desaturation, sleep fragmentation, and excessive daytime sleepiness (1, 2). OSA affects a significant proportion of the working population in many Western societies and is associated with increased rates of many chronic diseases (3, 4). Thus, untreated OSA may pose a significant public health problem (5). The recognition of OSA as a common clinical problem has intensified interest in identifying morphologic features that would both facilitate identification of high risk groups and be informative regarding underlying pathophysiology (6).

OSA appears to result from a variable combination of anatomic and pathophysiologic factors, some of which may be under genetic control (11). Obesity is the strongest risk factor for OSA in middle-aged adults (12). Increased neck circumference, which may be a marker for localized obesity, may also increase risk of OSA (14). Features of the upper airway and cranium also predispose to OSA (17). Many studies, using cephalometry or other imaging modalities almost exclusively, have described a number of possibly unique characteristics. These include a reduced posterior airway space (18), a long pharynx (7, 21), a longer and thicker soft palate (21), a long and large tongue (21, 24), a low position of the hyoid bone (21, 24), increased anterior lower facial height (25), deviation of head posture (26, 27), decreased sagittal dimension of the cranial base (17), and reduced anterior-posterior size of bony pharynx (17).

We recently reported that craniofacial risk factors for OSA, as measured by cephalometry, differed between African-Americans and white individuals (17). In whites, recognized risk factors included both bony and soft tissue measures: reduced anterior-posterior dimensions, reduced middle cranial fossa length, reduced intermaxillary length, increased tongue area, and increased soft palate length. In a sample of 41 African-Americans, these factors included only soft tissue measures: increased tongue area and increased soft palate length (28).

In the present study, we extended these observations by assessing the extent to which simple measures of head and facial form, made with anthropometric calipers, can be used to distinguish nonmorbidly obese subjects with and without OSA, and whether any relationship between these measures and OSA occurred in both whites and in African-Americans.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The study population consisted of 364 white and 165 African-American subjects who were participants in the Cleveland Family Study. Analyses were restricted to all participants who were both older than 25 yr of age and who had a body mass index (BMI)  32 kg/m². The age criterion was selected to minimize the effects of changes in craniofacial measures caused by growth. We excluded markedly obese subjects from these analyses to assess the effects of craniofacial structure in subjects without morbid obesity (a risk factor that might "overwhelm" the influence of structural risk factors on OSA susceptibility.) The overall cohort in this community-based genetic-epidemiologic study of OSA consists of "index" and control families, recruited using methods detailed previously (17, 11). Briefly, "index" families were identified through an index proband with laboratory-confirmed OSA. During the first 5 yr of the study, control families were also studied (later, the study focused on index families only, resulting in a predominance of index families). Control subjects were chosen randomly from a list of names provided by the index proband of neighbors/ friends who resided in the same neighborhood as the the index case and who had at least three living relatives available to study. Family members (n = 3,010) who were studied as part of the overall cohort included all available first-degree relatives and spouses, and selected second-degree relatives. Second-degree relatives were studied if they lived in the same household as members of the nuclear family or if the family previously had been found to have two or more relatives with OSA. The majority of subjects in this report consisted of subjects who were recruited as members of affected families (79% of the whites and 85% of the African-Americans).

Measurements

All subjects were studied in their homes, where they were visited by a trained technician who obtained anthropometric measurements, provided instruction on the use of an overnight sleep monitor, and provided instruction on the completion of questionnaires. All technicians underwent formal training and certification procedures regarding the study protocol, including anthropometric measurements (e.g., demonstrating < ± 2 mm differences in craniofacial and blood pressure measures among technicians). Every 3 to 6 mo, each technician was required to attend retraining sessions, and again demonstrate the ability to reliably obtain study measurements.

Anthropometric Data Collection

Anthropometric data included BMI, neck size, and head/facial form. Weight was measured on a digital scale, height was measured in stocking feet on a flat surface, and BMI was calculated as the weight (kg) divided by height squared (m²). Neck was measured at the level of the cricothyroid cartilage in a line perpendicular to the long axis of the neck using a nondistensible tape measure. Cranial and facial dimensions were measured by anthropometric spreading calipers (GPM; Seritex Inc., Carlstadt, NJ) to compute the cranial and facial indices. Cranial Index (CI) was defined as the ratio between maximum cranial width and maximum cranial length (maximum width × 100/maximum length) (28). Facial Index (FI) was defined as the ratio between nasion-gnathion height and bizygomatic width (nasion-gnathion height × 100/bizygomatic width) (28). Maximum cranial width is the greatest transverse diameter of the head, and maximum cranial length is the distance between the most inferior point below the glabella (the most prominent point, in the midsagittal plane between the eyebrows) and the farthest projecting point in the midsagittal plane, on the back of the head (opisthocranion) (28). Nasion-gnathion height is the distance between the point at which a horizontal tangential to the highest points on the superior palpebral sulci intersects the midsagittal plane [nasion] and the lowest median point on the lower border of the mandible [gnathion] (28). Bizygomatic width is the distance between the most laterally situated points on the zygomatic arches (28). (See AJRCCM Web site online data supplement at www.atsjournals.org for landmarks measured.) The long-term repeatability of the measures CI and FI, measured in 28 subjects at two time points 5 yr apart, showed correlations of 0.61 and 0.47, respectively.

Polysomnographic Data Collection

Polysomnography was performed with overnight monitoring using a portable sleep monitor (Edentech Models 1 or 2; Edentech Corp., Eden Prairie, MN) that measured and recorded nasal/oral thermistry, chest wall impedance, finger pulse oximetry, and heart rate. The apnea-hypopnea index (AHI) determined with this system has been found to correlate highly with the AHI from in-lab polysomnography (r = 0.96) (29). Apnea was defined as complete cessation of airflow accompanied by a 2.5% fall in oxygen saturation for at least 10 s. Hypopnea was defined as a discernible reduction in airflow compared with the baseline, associated with a 2.5% fall in oxygen saturation for at least 10 s. In all subjects, sleep time was estimated from visual inspection of the sleep record (i.e., periods of increased heart rate and movement artifact were considered "awake"), correlated with a sleep diary completed by the subject. The AHI was determined by dividing the number of apneas and hypopneas by the estimated hours of sleep time.

Statistical Analysis

All measures other than sex and race were treated as continuous variables. The AHI was transformed to its natural logarithm to achieve approximate normality. Means and standard deviations for each group (e.g., racial or OSA) were compared with a two-sample unpaired t test. The p values comparing groups (race or OSA) were calculated using Proc Mixed (SAS version 6.2; SAS Institute, Cary, NC), accounting for the familial clustering of data by adjusting for family as a random effect. Multiple linear regression analysis was used to analyze the relationship of the natural log-transformed AHI to CI and FI, adjusting for age, BMI, neck circumference, and sex. These analyses also were based on a random effects model, accounting for clustering by adjusting for family as a random effect (Proc Mixed). Separate models were created for whites and African-Americans. Subanalyses were restricted to participants with an AHI  15, an index commonly used to identify OSA of moderate or greater severity.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A majority of subjects in each racial group were recruited from index families (79% of whites and 85% of African-Americans). The percentage of male and female and African-Americans were proportionately comparable (46% of the whites and 45% of African-Americans were male). The distributions of the other demographic and anthropometric data are shown in Table 1. The mean neck circumference of the African-Americans was modestly but statistically significantly larger than that in the white sample. In the overall sample, there were insignificant differences between races in age, BMI, AHI, CI, and FI.

The variation of cranial and facial indices with OSA was examined within race-specific groups by comparing those with an AHI < 5 with those with an AHI  15 (Table 2). Among the whites, both cranial and facial indices showed significant differences between the OSA (AHI  15) and non-OSA (AHI < 5) groups. Specifically, in whites, OSA was associated with a higher CI (indicating a greater tendency towards brachycephaly) and a shorter facial height. In contrast, neither the CI or the FI varied with OSA among the African-Americans.

To assess whether the associations with cranial and facial indices differed between whites and African-Americans with OSA, analyses were restricted to participants with an AHI  15 (Table 3). Among these subjects with an AHI  15, both cranial and facial indices showed a significant difference between whites and African-Americans. The mean CI among whites with moderate-severe OSA was greater than the mean CI of African-Americans with OSA (82.3 ± 4.5 [SD] mm versus 79.9 ± 4.5 mm, p = 0.007). In contrast, African-Americans with OSA had a higher mean FI than that of the white OSA group (89.1 ± 9.0 mm versus 84.5 ± 6.1 mm, p = 0.004). In other words, whites with OSA had a greater tendency toward brachycephaly, a head form associated with reduced anterior-posterior cranial dimensions, than African-Americans with OSA. African-Americans with OSA, on the other hand, had a greater tendency to leptoprosopic facial types, i.e., a longer facial height and decreased facial width. These analyses were repeated after adjusting for age, sex, BMI, and neck circumference. As in the unadjusted analyses, the adjusted means showed that in those with OSA, whites had a higher CI than African Americans (82.3 ± 4.3 versus 79.8 ± 4.6; p = 0.010), and a lower FI (84.2 + 7.7 versus 89.4 + 7.3; p = 0.002).

To further assess the extent to which associations between AHI and the CI and FI were independent of age, sex, BMI, and neck circumference, we performed race-specific multiple linear regressions (Table 4). BMI and neck circumference were highly colinear (r = 0.52), and inclusion of both terms did not alter the significance of CI or FI as predictors of the AHI. Therefore, to improve interpretability, only the model with BMI, age, and sex adjustment are shown. After accounting for these variables and familial clustering, CI contributed significantly to the prediction in whites but not in African-Americans. After adjusting for CI, FI did not significantly predict AHI level in either racial group.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study, in which we extended previous observations on headform and OSA (17), has provided new data that characterize the association between anthropometric measures of cranial and facial form with the AHI in a large sample of whites and African-Americans. Previous studies of craniofacial risk factors for OSA have utilized predominantly cephalometry to study facial and upper airway dimensions. Differences in the size and the shape of the cranium and cranial base actually form the anatomic basis for subsequent variations in facial and upper airway morphology, however (30). For this reason, we thought that anthropometric measurements of head and facial form (CI and FI) would provide a useful appraisal of the determinants of anatomic variation.

In the classification of human head form, the two extremes are brachycephaly and dolichocephaly types (30). Each of these head form groups is associated with a characteristic facial form. Brachycephalic head forms are wider in the biparietal dimension and shorter in the anterior-posterior (occipital-frontal) dimension (CI  81.0). This head form is associated with a cranial base that is wider and shorter (30). Because the face is built on the cranial base, brachycephalic persons usually have wider and shorter (euryprosopic) facial forms. Dolichocephalic subjects exhibit a narrower biparietal width and relatively longer anterior-posterior length (CI  75.9), resulting in a cranial base that is also narrower and longer and a long-thin (leptoprosopic) facial form (30). Moreover, the cranial base forms the template for the airway, as well as the face. Thus, the dimensional, angular, and topographic characteristics of these structures vary according to head form type. In particular, the skeletal dimension of the pharynx is established by the size of the middle cranial fossathe area between the dorsum sellae and the center of the anterior border of the foramen magnum. The floor of this baso-cranial fossa is the roof of the pharyngeal compartment (30). A brachycephalic head form results in shorter anterior-posterior dimensions of the cranial base, greater cranial base flexure, reduced middle cranial fossa, and reduced upper airway anterior-posterior dimensions (30). Because the middle cranial fossa is shorter in these subjects, we hypothesize that this group should have greater anatomic risk of airway collapse. Consistent with this hypothesis, we have previously reported that a brachycephalic head form and a smaller mean posterior nasal spine to basion (PNS-Ba) distance were observed more frequently in the OSA members of sudden unexpected infant death families (31). Common anatomic risk factors may thus predispose to both sudden death in infancy and OSA. Other investigators (21, 32) have used lateral cephalograms to demonstrate that head form in patients with OSA tends toward brachycephaly by exhibiting a reduction in sagittal dimension of the cranial base.

The present study has provided corroboration and refinement of this hypothesis. The CI in white subjects with an AHI  15 is statistically significantly higher than that of non-OSA whites (Table 2). It is also significantly higher than that of African-Americans with OSA, in whom CI did not vary with AHI. This effect in whites is independent of age, sex, BMI, and neck circumference. Thus, brachycephaly is associated with an increased AHI in the white population but not in African-American population. Hard tissue anatomic risk associated with shape and size of the skeletal components of the face and upper pharynx may be relatively more important in whites. Other anatomic features (e.g., neck circumference) and soft tissue structures in the airway (e.g., tongue size) may be relatively more important risk factors in the African-American population.

We have previously reported a difference in the relationship of sleep-disordered breathing to age in whites and African-Americans (17). Mean AHI levels were higher for African-Americans at younger ages but drew closer to whites between ages 50 and 60, with a suggestion of lower levels at ages greater than 60 yr. The present findings provide additional evidence that the pathogenesis of OSA may differ in these racial groups.

Differences in the distribution of cranial and facial indices between whites and African-Americans were not significant for the whole study population (Table 1) but were significant for the group identified by an AHI  15 (Table 3). This finding suggests that there is considerable overlap in craniofacial morphology in Americans of European versus those of predominantly African descent. However, there appears to be a subgroup of European Americans, including those without morbid obesity, who appear to be especially predisposed to OSA because of a brachycephalic headform (higher CI). The CI, derived from simple measurements, may be useful for phenotyping and identifying population subsets with craniofacial risk factors for OSA.

Although FI differed between whites with and without OSA, FI did not predict the AHI after adjusting for age, BMI (and neck circumference), sex, and CI. This finding is consistent with the fact that the FI and CI are related to each other in a dependent fashion. Leptoprosopic facial types are more often associated with dolichocephalic head forms, and euryprosopic facial forms are more often associated with brachycephalic head forms. Although previous reports have suggested that facial form can be influenced by environmental factors (35) such as nasal obstruction, our data suggest that these two indices are not independent anatomic indicators of apnea risk. Because head form rather than facial form is more strongly associated with OSA, we suggest that the CI is probably a better anthropometric variable than FI to track for future epidemiologic studies of anatomic risk factors for OSA.

In summary, the findings in this study add to the limited cephalometric literature that suggests that the underlying anatomic substrate for OSA differs for whites and African-Americans. These data also provide support for the potential utility of anthropometric measurement of head form in studies of the OSA phenotype and for risk assessment. For some persons, head form may be an important factor in increasing susceptibility to OSA.