Airway and Cephalometric Changes in Adult Orthodontic Patients After Premolar Extractions

Increased attention is being paid to airway capacity, promoted in part by more widespread use of cone-beam computed tomography (CBCT) and concerns regarding obstructive sleep apnea.^1,–3  In addition, it seems there is a trend toward less frequent extraction and more arch expansion, especially in the mixed dentition.⁴  These trends are sometimes conjoined and supported by the belief that extraction of premolars contributes to collapse of the airway. In a survey of 125 orthodontists presented with a mouth-breathing female and a CBCT minimal cross-sectional area (MCA) reading of 50 mm², 42% said they would modify their treatment approach based on that information. The most common modification chosen (69%) was to avoid extraction.⁵ 

Previous studies investigating the effect of extractions on the airway have found either a reduction in dimensions after extractions^6,7  or no significant changes.^8,–10  It has been proposed¹¹  that the indication for extraction may be related to how the upper airway changes after treatment. For example, large retraction of the incisors to resolve bimaxillary protrusion after premolar extraction with maximum anchorage resulted in decreased upper airway volume,⁷  whereas airway dimensions increased following extractions as a result of severe crowding with minimum anchorage space closure.¹² 

Previous research has produced inconclusive findings. The purpose of this study was to investigate the relationship between extraction of premolars and changes in airway dimensions and possible associated skeletal and dental features in a sample of nongrowing patients. This research was undertaken to address four questions:

1. Is extraction as part of treatment associated with changes in airway dimensions?

2. Is extraction as part of treatment associated with selected skeletal and dental dimensions?

3. Do initial skeletal and dental dimensions interact with extraction to change airway dimensions?

4. Do measurement features of CBCT mask changes in airway, skeletal, and dental features?

Sample selection began with identifying all patients who were treated between January 2007 and June 2018 in the Orthodontics Department at the University of the Pacific, Arthur A. Dugoni School of Dentistry, with appropriate CBCT images obtained as part of routine diagnostic records. Subjects met the following inclusion criteria: they were at least 18 years old at initiation of orthodontic treatment and there was availability of pre- (T1) and posttreatment (T2) CBCTs. After exclusion criteria were applied (Table 1), 221 patients were eligible to participate in this study (Figure 1). To be included in the extraction group, patients had to have had at least two premolars extracted. This condition was determined by the fact that shortening of the arch dimension by incisor retraction is primarily achieved by premolar extraction. Subjects in the nonextraction group were selected by matching for age and sex with subjects in the extraction group (Table 2). This sampling process resulted in 42 matched subjects for the extraction and nonextraction groups. One extraction subject was later excluded as a result of poor CBCT scan quality. Therefore, the final sample of 83 subjects included 41 subjects in the extraction group and 42 subjects in the nonextraction group. Within the extraction group, six subjects had only upper premolars extracted, and 35 subjects had four premolars extracted.

This study was approved by the institutional review board at the University of the Pacific (18-27).

Seventy-three CBCT scans were taken using the i-CAT Classic unit with 40-second exposure time (23 cm × 17 cm FOV, 0.3-mm voxel size), and 93 CBCT scans were taken using the i-CAT Next Generation unit with 8.9-second exposure time (23 cm × 17 cm FOV, 0.3-mm voxel size). While taking the scans, subjects were sitting upright and were instructed to bite into maximal intercuspation and remain stationary without swallowing. No instructions were given regarding the mode of breathing or tongue position. DICOM images were imported into Invivo software (version 6; Anatomage, San Jose, Calif) and deidentified for analysis. All scans were oriented as follows: (1) axial plane was adjusted by aligning the inferior border of the right and left orbits; (2) coronal plane was determined by the Frankfort horizontal plane (right porion and right orbitale); and (3) sagittal plane was adjusted by aligning the most anterior point on the lateral borders of the right and left orbital rims. Skeletal and dental landmarks were identified using the 3D Analysis function of Invivo software (Table 3).

For purposes of analysis, the airway was divided into three regions: nasopharynx (superior to the palatal plane), retropalatal region of the oropharynx (between the palatal plane and the base of the soft palate at the most anterior point), and retroglossal region of the oropharynx (between the base of the soft palate and the base of the epiglottis). Airway segmentation threshold values were adjusted to eliminate imaging artifacts and were held constant at about 400 relative Hounsfield units. The airway volume was then calculated in cubic centimeters, and the most constricted MCA of the airway was calculated in square millimeters for each of the three regions using the airway function of Invivo software. Additionally, in order to standardize measurements based on anatomical landmarks, the sagittal length (AP, in millimeters), lateral width (Lat, in millimeters), and cross-sectional area (XSec, in square millimeters) were measured at the level of the palatal plane, the base of the soft palate, and the base of the epiglottis (Figure 2; Table 4). In total, 15 airway and 10 cephalometric measurements were determined by averaging the measurements from two judges (Figures 2 and 3). Table 4 summarizes the variables used in this study.

All airway measurements were performed by two judges. The average of their values was used for analysis, and the consistency of their reading of the CBCT images was gauged by Cronbach's alpha, a universal ICC measure. Descriptive statistics were calculated for 15 airway and 10 skeletal and dental variables at T1 and T2.

Independent t-tests were performed to determine whether groups were comparable at baseline. Repeated-measures analysis of variance (ANOVA) tests were performed to test for differences across time and across group (extraction vs nonextraction) and to identify interactions showing that extraction differentially affected airway, skeletal, or dental features. Repeated-measures tests were appropriate given that failure to consider within-subject variance typically overestimates the sensitivity of tests. Correlations between baseline scores and change in score, using Pearson correlation coefficients, were performed to identify potential confounding measurement sources of variance. Sequential-entry multiple regression analysis was performed to control for potentially confounding effects. Exact test statistics are reported, as are exact P-values for all tests.

The average Cronbach alpha coefficient for consistency in airway readings for two judges was 0.981, with no coefficient below 0.954.

The results for airway measures are shown in Table 5. The mean and standard deviation 2 × 2 crossed classification of scores for the five airway measures in the nasopharynx, retropalatal, and retroglossal regions are shown. Figure 4A is a graph of a representative result for retropalatal volume. This is typical of the findings by virtue of showing no change over time, a slight tendency for extraction cases to have larger measurements both initially and at the end of treatment, and no differential effects with extraction.

Although subjects were matched based on age and sex, for seven of the 15 airway features measured, the extraction group began the study with slightly larger airways (independent t-tests for differences between groups at baseline). In five of 15 cases, subjects in the extraction group showed larger airway values when combined across both T1 and T2 (main effect for group in repeated-measures ANOVA). There was only one feature (MCA in the retropalatal areas) for which that effect showed a marginally significant difference during treatment, regardless of extraction (main effect for time in repeated-measures ANOVA). It appeared that there was a differential effect of extraction in the single case of MCA in the nasopharynx region. That difference failed to reach significance at P < .10 when corrections are taken for measurement error.

The only consistent finding in these data was that changes in airway features were associated with their baseline conditions. For all but one feature, the correlation between baseline score and change score was highly significant. The association was always negative, demonstrating that small airways grew in size and large airways decreased in size. This effect is diagrammed in Figure 4B for retropalatal MCA, showing that both extraction and nonextraction cases that were initially smaller than 100 mm² tended to increase and that both extraction and nonextraction cases that were larger than 200 mm² tended to decrease in size during treatment.

Table 6 is a parallel summary of the analysis of three skeletal and seven dental features as they were affected by extraction or nonextraction. The positions of both the maxillary and mandibular incisors were larger in the extraction group initially but decreased significantly in the extraction group and increased significantly in the nonextraction group. The same significant interaction was observed for both the maxillary and mandibular intermolar widths. The opposite effect was seen with respect to interincisor angle, which indicates the impact of extraction on incisor inclination. Very large associations were observed between baseline and change measures for five of the dental features representing incisor inclination and intermolar width.

Correlation matrices were calculated relating changes in the 15 airway features with both baseline and change measures for the 10 skeletal and dental features. These calculations were performed for both the total sample and separately for only the extraction subjects. Of the 600 coefficient values calculated, only seven of them were statistically significant.

A retrospective analysis was conducted on 83 nongrowing, orthodontically treated patients, matched for sex and age, in which half received premolar extractions and half did not. Five features in each of the nasopharynx, retropalatal, and retroglossal airway regions and 10 skeletal and dental features were assessed at the beginning and end of treatment. Results were analyzed across time, across treatment type, and for interactions. Associations were also explored between airway and skeletal and dental measures.

The results supported the conclusion that extraction has an effect on dental features such as upper and lower incisor position and inclinations and intermolar width. However, there was no evidence that extraction changed sagittal and transverse distances or minimal cross-sectional area or volume in the nasopharyngeal, retropalatal, or retroglossal regions. In addition, there was no evidence that changes in the measured skeletal or dental features had an indirect effect on airway features. Strong evidence emerged that common measurement effects in repeated measures may obscure the types of conclusions drawn from similar clinical cases.

Two early studies^6,7  that reported a negative impact of extraction on airway did not use CBCT technology, so they could not report transverse dimensions, area, or volume. They reported correlations on changes rather than differences across groups and did not use a control group. Pliska et al.¹⁰  reported results similar to those found in the current study. Their sample contained patients with larger initial airways and very large standard deviations,¹⁰  but their results were similar in finding regression toward mean values. With respect to skeletal and dental features, the findings in the current study were similar to those reported previously in the literature.^9,12,–14 

Averaging measurements across subjects masked conclusions about both individual and aggregate changes in airway features. This was the largest and most consistent finding in this research and an effect that applied equally to measures of airway and skeletal and dental features. In 19 of the 25 variables analyzed, this effect was significant at P < .01. The consistent negative associations between baseline score and change indicated that small values at T1 were more apt to grow in size by T2 and that initially large values consistently got smaller, which creates a chance for bias, especially if only large initial airways are studied. Such sampling restrictions are sometimes used on ethical grounds to mitigate supposed risk to patients with small airways or, in general, if adequately large airways dominate a sample.

Often, the difference between repeated measures of a feature can be attributed to treatment changes, but it is not always possible to isolate these effects from confounding factors such as measurement artifacts and natural growth. In most cases, the effects of measurement and growth are not random, and they interact with planned interventions.

Measurement error across repeated measures is known as “regression toward the mean.”¹⁵  Values far from the mean at T1 are seen to be closer to the mean at T2 because of the underlying nature of random sampling. It is unlikely that extreme values will become even more extreme. The more unreliable the measurement system, the larger the regression toward the mean. This effect can be approximated by looking for negative correlations between baseline score and change from T1 to T2 or by calculating the ICC with the difference between individual scores at baseline and baseline average as a covariate. This study showed an extremely high consistency in the reading of CBCT images of airway features but significant and consistent inconsistency in the taking of these images. Postural and functional fluctuations from one exposure to the next can vary depending on the type of feature being measured.

The conclusions of this study can only be generalized to nongrowing patients and are limited to conclusions measured by the variables chosen. A large number of zero-order correlations were calculated between skeletal and dental features and characteristics of the airway, and they were found to be insignificant. However, it remains possible that patterns of several skeletal and dental features might have an effect. Multiple regression and latent structure analyses may prove useful in clarifying this relationship. It was clear that factors such as planned therapeutic effects, natural growth, and the interaction of anatomical features are confounded with the variable timing of growth across individuals and the statistical complications of using measurement systems with significant potential for unreliability across repeated measures. Much work remains to be done to disentangle these sources of variation in order to prevent mistaking the reasons for observed changes or, in this case, the absence of changes, averaging across patients.

The assessment of airway dimensions utilizing CBCTs is subject to many limitations. For example, body mass index was not recorded, which might affect the airway dimensions. In addition, patient positioning and the process of breathing have been shown^16,,–19  to change upper airway volume, size, and shape. Patients with narrow upper airway cross sections can maintain airway patency by dilating their airway,²⁰  illustrating that the dimensions of the airway are dynamic and variable.

• Although the indications for extractions in orthodontics have been debated, the fact remains that some cases necessitate removal of teeth in order to achieve treatment and/or esthetic goals. As a result of the variability in airway dimensions, it is advised to consider additional diagnostic information, such as clinical examination, sleep questionnaires, and polysomnography, rather than relying solely on airway measurements from a CBCT scan, when making treatment decisions regarding airway concerns.

• Claims that extraction treatment is generally contraindicated because of airway considerations are too general to be useful. Extraction cases should be evaluated on a case-by-case basis, preferably with the aid of additional diagnostic evaluations, such as an obstructive sleep apnea screening. Characteristics of the airway at baseline must be taken into account, as must the possibility of wide variation due to the unreliability of CBCT imaging in evaluating certain features. Due partly to patient selection, natural growth, and measurement effects, small airway dimensions at baseline will appear to increase, and large airway dimensions will appear to decrease after treatment.