Control patient airway measures (5-month-old): A, airway divisions; B, tongue size and position; C, hyoid position; D, lateral mandible measures; E, inferior pogonial angle; F, bigonial distance.
A, Larger tongue area, taller tongue height relative to the vallecula, and more posterior tongue position are shown in RS compared with control consistent with glossoptosis. B, Compressed hyoid position is illustrated in patient with RS compared with control. C and D, A shorter and flatter mandible is demonstrated in a patient with RS compared with control.
A, Boxplot of 10-measure (FDR <0.2, RPD <10%) and 5-measure simplified CT composite scores for tracheotomy vs no tracheotomy . B, ROC curves for all-measure, 10-measure (FDR <0.2, RPD <10%), and 5-measure simplified CT composite scores, with sensitivity and specificity indicated in parentheses. FDR indicates false discovery rate; ROC, receiver operating characteristic; RPD, relative percent difference.
eTable. Calculation of Patient’s Simplified Composite Score for Predicting Tracheotomy Risk.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Lee VS, Evans KN, Perez FA, Oron AP, Perkins JA. Upper Airway Computed Tomography Measures and Receipt of Tracheotomy in Infants With Robin Sequence . JAMA Otolaryngol Head Neck Surg. 2016;142(8):750–757. doi:10.1001/jamaoto.2016.1010
Airway management in infants with Robin sequence is challenging. Objective upper airway measures associated with severe airway compromise requiring tracheotomy are needed to guide decision making.
To define objective upper airway measures in infants with Robin sequence from craniofacial computed tomography (CT) and to identify those measures in Robin sequence associated with tracheotomy.
Design, Setting, and Participants
A cohort study (2003 to 2014, over 1-year follow-up) of 37 infants with Robin sequence evaluated for surgical management and 37 selected age- and sex-matched controls without a craniofacial condition conducted in a pediatric institution’s craniofacial center.
Main Outcomes and Measures
Define and compare CT-generated upper airway measures in these groups: infants with Robin sequence vs controls, and infants with Robin sequence with vs without tracheotomy. A negative difference signifies lower values for the Robin sequence and tracheotomy groups. Clinical data collected included age and height at time of CT scan, sex, tracheotomy presence, associated syndrome, and laboratory indicators of hypoventilation and hypoxemia. To evaluate interrater reliability, 2 raters performed each measurement in the Robin sequence group.
In 74 infants, 17 of 28 measures were different between infants with Robin sequence and those in the control group. Tracheotomy was performed in 14 of 37 (38%) infants with Robin sequence. Infants with tracheotomy more commonly had associated syndromes (12 of 14 [86%] vs 11 of 23 [48%]) and a history of hypoventilation and hypoxemia (13 of 14 [93%] vs 15 of 23 [65%]). Five of the 11 measures associated with tracheotomy were reliable and simpler to measure with the following mean differences (95% CIs) between groups: tongue length, 0.87 (0.26 to 1.48); tongue position relative to palate, 0.83 (0.22 to 1.45); mandibular total length, −0.8 (−1.42 to −0.19); gonial angle, 0.71 (0.08 to 1.34); and inferior pogonial angle, 0.66 (0.02 to 1.29). Using a receiver operating characteristic analysis, a composite score of these 5 measures for predicting tracheotomy risk yielded an area under the curve of 0.83 and achieved 86% sensitivity and 74% specificity.
Conclusions and Relevance
Computed tomography measures quantifying tongue position and mandibular configuration can identify infants with Robin sequence, and importantly, differentiate those who have severe upper airway compromise requiring tracheotomy. Following validation, these measures can be used for objective upper airway assessment and for expediting clinical decision-making in these challenging cases for which no such tools currently exist.
Robin sequence (RS), classically defined as the triad of micrognathia, glossoptosis, and upper airway obstruction, commonly with cleft palate (although whether this is an obligatory feature of RS is controversial), affects 1 in 8500 to 20 000 neonates, and may be associated with underlying anomalies or syndromes.1-4 Infants with RS present with upper airway obstruction that traditionally has been thought to be localized to the tongue base owing to glossoptosis.5,6 The severity of upper airway obstruction varies, ranging from snoring to life-threatening airway compromise requiring tracheotomy.3,7
Options to stabilize the airway in infants with RS include side and prone positioning, nasopharyngeal airway placement, and mandibular distraction.5 In our institution, tracheotomy is pursued as a last-resort and definitive airway treatment for patients with the most severe upper airway obstruction. Long-term tracheotomy dependence impacts morbidity and mortality, including risk of death, infections, bleeding, tracheal stenosis, and speech and swallowing difficulties. There are no evidence-based tools or systematic algorithms to guide airway management decisions in patients with RS.8 Treatment based on anecdotal experience varies within and among institutions and may cause harm and lead to prolonged hospital stays or multiple operative procedures.
Computed tomographic (CT) imaging can be used to evaluate bony and soft tissue anatomy of the upper airway of the infant in 2 and 3 dimensions, which is not possible with cephalometrics.9-17 Prior systematic RS imaging assessments have focused on changes in facial skeletal morphology with mandibular distraction, not soft tissue morphology, which is crucial for airway improvement.18-24 Since craniofacial CT is routinely used to determine the location of upper airway obstruction and mandibular anatomy in infants with RS, for whom surgical intervention is being considered, we sought to increase the information that we derive from these scans. Our multidisciplinary research team developed a comprehensive list of previously validated as well as novel CT measures characterizing volume, area, length, and angles of the upper airway. The purpose of our study was to establish the reliability of these measures, and to identify those measures in patients with RS that differ from those in non-RS patients and are associated with tracheotomy. We hypothesized that a subset of CT measures would differentiate treatment course in patients with RS, specifically whether or not patients had a tracheotomy.
Question What upper airway computed tomography (CT) measures in infants with Robin sequence are associated with receiving tracheotomy?
Findings In this case-centered study of 74 infants, 5 of the 11 measures associated with receiving tracheotomy were reliable and simpler to measure: tongue length, tongue position relative to the palate, mandibular total length, gonial angle, and inferior pogonial angle. A composite score of these 5 measures was reliable for predicting tracheotomy risk.
Meaning Upper airway CT measures quantifying tongue position and mandibular configuration can be used to differentiate infants with severe upper airway compromise requiring tracheotomy.
This study was approved by the Seattle Children's Hospital (SCH) institutional review board and need for consent for retrospective review was waived. The RS group was identified from a longitudinal database of patients seen at our craniofacial center from 2003 to 2014. Potential patients had a diagnosis of RS and/or 2 or more of the following: micrognathia, glossoptosis, cleft palate, and airway obstruction (n = 137). Our multidisciplinary team followed a comprehensive algorithm using physical examination, laboratory, endoscopic, and polysomnography findings to assess the severity of airway obstruction. Medical followed by surgical interventions were pursued to stabilize the airway. Craniofacial CT scans with 3 dimensional (3D) reconstruction were prospectively obtained in infants being considered for presurgical distraction. Patients were included if they had a premandibular distraction CT scan available regardless of whether medical interventions or tracheotomy had been performed.
The control group was identified from the SCH database of existing CT data, the picture archiving and communication system (PACS), which includes all SCH clinical imaging data. Participants in the control group were age- and sex-matched to those in the RS group and selected based on age-matching parameters (within 2 weeks for infants <3 months of age, within 1 month for infants 3-12 months of age, and within 3 months for children >12 months of age). Controls were excluded if they had any craniofacial condition or prior facial or airway surgery, including tracheotomy. Computed tomographic scans included the face and upper airway above the level of a plane parallel to the Frankfort horizontal plane passing through the anterior inferior aspect of C4.
Among infants with RS, descriptive data were collected, including demographics and history (age at time of CT scan, sex, birth weight, gestational age, associated syndrome, and reason for CT scan), examination and laboratory details (height at time of CT scan, presence of secondary cleft palate, serum carbon dioxide, and oxygen saturation levels), and polysomnography measures (obstructive apnea-hypopnea index). Outcome data were collected, including whether medical intervention (positioning or nasopharyngeal airway placement), mandibular distraction, or tracheotomy was performed, and the age at intervention. Among infants in the control group, age and height at time of CT scan, sex, and reason for CT scan were ascertained.
Computed tomographic scans were obtained as part of clinical management using standard institutional protocols with slice thickness between 0.625 mm and 1.25 mm and patients scanned in the supine position. Axial images were reformatted parallel to the Frankfort horizontal plane, and coronal and sagittal images were subsequently generated, providing a standardized reference plane. An experienced rater (V.S.L.) performed CT analysis for both the RS and control groups. An additional rater (either K.N.E. or F.A.P.) performed a second reading for the RS group to evaluate interrater reliability. Raters performed CT analysis in a random order with respect to age, and without knowledge of clinical severity of airway compromise or outcome. Image processing software (OsiriX; Pixmeo) was used to perform all CT reformats and analysis.25
The airway was divided into the nasopharynx, oropharynx, and hypopharynx (Figure 1A). Airway volumes for each division were calculated on axial images using region-of-interest (ROI) analysis, set at a threshold for air density, and the ROI volume calculator. Volumes occupied by the radiopaque border of an artificial airway were not included in the reported nasopharynx, oropharynx, and hypopharynx volumes. Maximum cross-sectional areas and cross-sectional areas at the inferior border for each division were calculated on the same axial series as the volumes. Craniocaudal lengths for each division were calculated on the reformatted sagittal images. Next, sagittal midline measures related to tongue dimensions and positioning and hyoid positioning were performed (Figure 1B and C). Finally, mandible measures were performed using 3D reconstructed lateral, inferior, and anterior views (Figure 1D-F). Detailed definitions for measures are provided in the eMethods in the Supplement.
All measures except for angles were normalized to each patient’s skull size, according to the formula:
, where y is the measure, d is the measure’s dimensionality (length = 1, area = 2, volume = 3), and yNB is the patient’s nasion to basion distance.26,27
To assess interrater reliability, for each pair of readings on a single measure and patient with RS, relative percent difference (RPD), ie, the difference divided by the mean and expressed in percent, was calculated. For each measure, the mean RPD across all patients with RS was reported.
For comparisons (RS vs control and RS with vs without tracheotomy), a paired t test and linear regression, respectively, were used to test for differences in measures between the groups. Sensitivity analyses evaluated the effect of age adjustment. To make differences more comparable, the mean difference in a measure between groups is presented in units of that measure’s standard deviation for the RS group. A negative mean difference signifies lower values for the RS and tracheotomy groups. For example, a −1 mean difference in mandibular body width between patients in the RS group and those in the control group indicates that the mean mandibular body width was shorter for those in the RS group, by 1 RS group’s standard deviation. Multiple hypothesis testing was corrected for using false discovery rate (FDR) methodology.28 An FDR less than 0.2 was considered significant, and 95% CIs and P values were calculated.
To evaluate strategies for combining CT measures for predicting tracheotomy risk in the RS group, leave-one-out cross-validation was performed. For each patient, each normalized measurement was standardized by the mean and standard deviation of all other patients. In addition, the direction of the difference between tracheotomy and no tracheotomy groups for each measure (still excluding said patient) was noted as +1 or −1. For a score composed of k measures, the standardized measurements participating in the score were multiplied by the direction, and the result summed and divided by k, producing a personal average tracheotomy risk score for each patient. For each strategy, a receiver operating characteristic (ROC) curve was calculated by associating the leave-one-out tracheotomy risk scores for all patients with the binary endpoint of whether tracheotomy was performed.
Descriptive data are summarized in Table 1. A total of 37 age- and sex-matched infants with RS and 37 controls were included. There was 1 sex mismatch (female infant with RS, male control) owing to the limited number of eligible control CT scans performed on infants at our institution. Tracheotomy was performed in 38% (14/37) of infants with RS at a median age of 2.7 (range, 0.1-69.0) months and prior to CT scan in 16% (6/37) of infants with RS. Infants with tracheotomy were more commonly female (11 of 14 [79%] vs 15 of 23 [65%]), had associated syndromes (12 of 14 [86%] vs 11 of 23 [48%]), and had a history of hypoventilation and hypoxemia (13 of 14 [93%] vs 15 of 23 [65%]).
Mean RPDs for the measures ranged from 2% to 19% (median, 7%; interquartile range, 5%- 11%; Table 2). Our research team agreed that an RPD of 10% or less represents an acceptable level of reproducibility by another image rater.
Mean differences between groups, 95% CIs, and P values are presented in Table 2. Comparing patients in the RS group with those in the control group, 9 measures were significantly larger in patients in the RS group (volume and cross-sectional area at the inferior border of the hypopharynx; craniocaudal length of the oropharynx; tongue area, height, and position relative to the anterior nasal spine; hyoid craniocaudal position; and gonial and inferior pogonial angles), and 8 measures were significantly smaller in patients in the RS group (maximum cross-sectional area and cross-sectional area at the inferior border of the nasopharynx; hyoid posterior distance to the vertebrae and anterior distance to the mandible; and mandibular ramus height, body length, total length, and body width). Significant measures are demonstrated in a representative 2-week-old patient with RS and matched control (Figure 2). Sensitivity analyses adjusting for age identified the same measures.
Within the RS group, 11 measures were associated with receiving tracheotomy: smaller volume of the oropharynx, smaller maximum cross-sectional area of the oropharynx, larger tongue area, longer tongue length, higher tongue position relative to the palate, shorter hyoid anterior distance to the mandible, shorter mandibular ramus height, shorter mandibular body length, shorter mandibular total length, flatter gonial angle, and flatter inferior pogonial angle. Sensitivity analyses adjusting for age identified the same measures.
For predicting tracheotomy risk for those in the RS group, a composite score was generated incorporating the 10 measures that were both associated with tracheotomy (FDR <0.2) and reliable (RPD <10%) (includes the mentioned 11 measures except hyoid anterior distance to the mandible). In cross-validation analysis, this composite score differentiated between the tracheotomy (median, 0.3; interquartile range, 0.1-0.5) and no tracheotomy (median, −0.4; interquartile range, −0.8 to 0.1) groups (Figure 3A). The FDR <0.2 and RPD <10% composite score for predicting tracheotomy risk yielded an area under the ROC curve of 0.81. Using a threshold value of >0 for predicting tracheotomy risk, the FDR<0.2 and RPD <10% composite score had 79% sensitivity and 65% specificity (Figure 3B).
In an attempt to simplify the CT composite score for routine clinical use, a composite score was generated incorporating those 5 measures that were associated with tracheotomy (FDR <0.2), reliable (RPD <10%), readily obtained without significant CT postprocessing steps (ie, excluded volume and area measures), and unique (excluded mandibular ramus height and body length, which correlated with mandibular total length): tongue length, tongue position relative to the palate, mandibular total length, gonial angle, and inferior pogonial angle. This simplified composite score differentiated between the tracheotomy (median, 0.3; interquartile range, 0.03-0.7) and no tracheotomy (median, −0.4; range, −0.6 to 0.1) groups (Figure 3A). The simplified composite score for predicting tracheotomy risk yielded an area under the ROC curve of 0.83. Using a threshold value of >0 for predicting tracheotomy risk, the simplified composite score had 86% sensitivity and 74% specificity (Figure 3B). Detailed instructions and template on how each patient’s simplified composite score was calculated are provided in the eResults and eTable in the Supplement.
Using objective measures from standard, clinically obtained CT scans, differences in upper airway morphology between infants with RS and controls can be characterized. Compared with controls, infants with RS had a larger tongue area, taller tongue height relative to the vallecula, and more posterior tongue position, quantifying the degree of glossoptosis (Figure 2A). In addition to shorter and flatter mandibles, infants with RS had a lower hyoid position relative to the palate and hyoid position compressed between the mandible anteriorly and vertebrae posteriorly, illustrating the external (mandible) and internal (hyoid) bony relationships that influence oral and pharyngeal soft tissue shape and easily narrow the compliant airway of those with RS to the point of obstruction (Figure 2B-D). Unlike prior studies, our study compares upper airway morphology in infants with RS with controls using objective imaging data. It confirms that the shorter mandibular dimensions and flatter gonial angles present in older patients with RS are also present in infants.9,12,15,18,22,24
Using pretreatment CT scans, we also identified differences in upper airway morphology in infants with RS treated with tracheotomy compared with those treated with positioning, nasopharyngeal airway placement, or mandibular distraction. Infants with RS receiving tracheotomy had a smaller oropharyngeal volume and cross-sectional area, larger tongue area, longer tongue length, and higher tongue position relative to the palate. These findings demonstrate quantitatively more severe glossoptosis in the tracheotomy group, and that the tongue in these patients occludes both the nose and oropharyngeal inlet. The tracheotomy group also had shorter mandibular dimensions and flatter mandibles.9,12,15,18,22,24 This study links pretreatment airway measures to treatment outcomes.
The airway measure findings in those in the RS group compared with those in the control group, specifically the smaller cross-sectional area of the nasopharynx, longer craniocaudal length of the oropharynx, and larger volume and cross-sectional area of the hypopharynx, may indicate dilation above and below the level of maximal narrowing, the oropharynx. They also, however, highlight the dynamic nature of the airway and illustrate the limitations of 3D airway measures, which capture a single snapshot in time and are influenced by factors such as respiratory phase that are impossible to control or standardize in infants with respiratory compromise. We acknowledge that CT scans represent static images of the airway and can be combined with dynamic assessments. An objective upper airway assessment tool optimally should incorporate CT measures that focus on well-defined bony and soft tissue relationships and quantify the tongue position and mandibular configuration characterizing patients with RS. Our simplified composite score consists of 5 easily obtained length and angle measures (tongue length, tongue position relative to the palate, mandibular total length, gonial angle, and inferior pogonial angle) and was most sensitive and specific for predicting tracheotomy risk in infants with RS. Our promising results support the need for further studies designed to evaluate the efficacy and utility of incorporating imaging-derived measurements into airway obstruction stratification and decision-making tools for infants with RS.
This is an exploratory study and has limitations. The small sample size limits the power of our study and the robustness of our conclusions. Relative to the published literature, however, this is the youngest and largest RS group with objective airway morphology data, and the first study to compare infant airway and facial skeletal morphology with a control group. Furthermore, the potential for selection bias exists, because all CT scans were obtained for clinical indications in the RS and control groups. Computed tomographic scans were performed in infants with RS with more severe airway obstruction, and whether the morphological differences would persist in a population of infants with milder involvement is unknown. In addition, heterogeneity was present within each group, but any attempts to reduce this would have further compromised the small sample size. We attempted to minimize confounding in the RS vs control comparison by matching based on age and sex, arguably the covariates with the most potential for confounding airway and skeletal measurements.3,10,12 The normalization by nasion to basion distance served to further address confounding by the size of the patient.26,27,29 There were a higher proportion of syndromic infants in the tracheotomy group, which may be related to the severity and levels of airway obstruction or other comorbidities. Future studies will evaluate specific imaging differences within more homogeneous subgroups. Moreover, there is the potential for measurement bias. The CT measures evaluated in this study are defined by bony or soft tissue landmarks, but differences may be introduced by human variation in determining the exact location of these landmarks. We, however, expect these differences to be small given the high reliability observed between raters in this study. Computed tomographic scans in patients with RS also more commonly contained airway devices and enteric tubes, which may have further reduced airway volumes that were calculated from air density on CT scan and therefore did not include the margins of the tube. Although we were unable to estimate these differences using the available imaging software, we predict them to be minimal. To further address this, length and angle measures, which are less likely to be impacted by the presence of devices and tubes, were included in the composite score. Standard cephalometric mouth closed position was also not possible for these clinical scans obtained in infants with respiratory compromise. To minimize the impact of mouth posture and positioning that may also have been a source of measurement bias, our skeletal measures focused on evaluating the dimensions of the mandible. Future work should also investigate maxillary involvement, maxillomandibular relationships, and symphyseal plane rotation, particularly when assessing changes with mandibular distraction.19 Finally, we chose the outcome of treatment with tracheotomy to represent the presence of more severe upper airway obstruction in infants with RS, and we acknowledge this assumption may be subject to bias. The presence of a tracheotomy at the time of CT scan in a small subset of patients in our RS cohort may also have influenced our measures, and the effect of airway treatment on the airway phenotype will be a focus of future work.
This work represents a first step toward development of an evidence-based decision tool for airway treatment in patients with RS, but prospective validation is needed. To enrich our understanding of the severity of airway obstruction in infants with RS, we also plan to examine the validity of the airway and skeletal measures compared to relevant outcomes and clinical characteristics, including polysomnography, airway endoscopy, and other measures of clinical severity. Future work could also evaluate the validity of these imaging measures using other modalities without ionizing radiation, such as magnetic resonance imaging.
Computed tomographic measures may objectively and quantifiably evaluate glossoptosis and mandibular underdevelopment in patients with RS and thereby stratify severity of upper airway obstruction. Ultimately, if validated in a larger series, these measures could be incorporated into infant airway assessment tools to guide decision making.
Corresponding Author: Victoria S. Lee, MD, Department of Otolaryngology–Head and Neck Surgery, University of Washington, PO Box 356515, Seattle, WA 98195 (firstname.lastname@example.org).
Published Online: June 2, 2016. doi:10.1001/jamaoto.2016.1010.
Author Contributions: Drs Lee and Perkins had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Evans, Perez, Oron, Perkins.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Lee, Evans, Oron.
Critical revision of the manuscript for important intellectual content: Evans, Perez, Oron, Perkins.
Statistical analysis: Oron, Perkins.
Obtained funding: Lee, Evans, Oron, Perkins.
Administrative, technical, or material support: Evans, Perez, Perkins.
Study supervision: Lee, Evans, Perez, Perkins.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported by a Centralized Otolaryngology Research Efforts American Society of Pediatric Otolaryngology Research Grant, award number 349553 (Dr Lee), Seattle Children’s Center for Clinical and Translational Research Scholars Program (Dr Evans), and the National Center for Advancing Translational Sciences, award number UL1TR000423 (Dr Oron).
Role of the Funder/Sponsor: The funding sources had no role in the design or conduct of the study; collection, management, analysis, or interpretation of data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of any of the funding sources.
Previous Presentation: This study was presented as an oral presentation at the American Society of Pediatric Otolaryngology Combined Otolaryngology Spring Meetings; April 24, 2015; Boston, Massachusetts.
Additional Contributions: Erik Stuhaug, AAS, from the Seattle Children's Craniofacial Center assisted with the formatting of the figures for this manuscript. He was compensated for his contribution.
Create a personal account or sign in to: