Assessment of dentofacial growth deviation in juvenile idiopathic arthritis: Reliability and validity of three-dimensional morphometric measures

Introduction Patients with juvenile idiopathic arthritis (JIA) and involvement of the temporomandibular joint (TMJ) often experience abnormal facial growth. Three-dimensional (3D) assessment of dentofacial growth deviation has become more common with advancement and commercialization of imaging technologies. However, no standardized guidelines exist for interpretation of 3D imaging in patients with JIA. The aim of this study was to propose and validate morphometric measures for the 3D radiographic assessment of dentofacial growth deviation in patients with JIA to enhance: 1) Description of dentofacial growth deviation; 2) Treatment planning; 3) Longitudinal follow-up. Methods The study was conducted in a standardized sequential-phased approach involving: 1) Preliminary decision-making; 2) Item generation; 3) Test of content-validity; 4) Test of reliability; 5) Test of construct validity; 6) Establishment of final recommendations. Results Twenty-one morphometric measures were evaluated. Based on results of reliability and validity-testing including subjects with JIA (n = 70) and non-JIA controls (n = 19), seven measures received a “high recommendation” score. Those measures were associated with posterior mandibular height, occlusal cant, mandibular asymmetry, mandibular inclination, and anterior/posterior lower face height. Nine other measures were “moderately recommended” and five received a “somewhat recommendation” score. Conclusion Seven morphometric measures were considered very useful in the 3D assessment of growth deviation in patients with TMJ disease associated with JIA. These variables can be used to standardize the description of dentofacial deformities and to plan corrective interventions.


Introduction
Patients with juvenile idiopathic arthritis (JIA) and involvement of the temporomandibular joint (TMJ) often experience abnormal facial growth. Three-dimensional (3D) assessment of dentofacial growth deviation has become more common with advancement and commercialization of imaging technologies. However, no standardized guidelines exist for interpretation of 3D imaging in patients with JIA. The aim of this study was to propose and validate morphometric measures for the 3D radiographic assessment of dentofacial growth deviation in patients with JIA to enhance: 1) Description of dentofacial growth deviation; 2) Treatment planning; 3) Longitudinal follow-up.

Methods
The study was conducted in a standardized sequential-phased approach involving: 1) Preliminary decision-making; 2) Item generation; 3) Test of content-validity; 4) Test of reliability; 5) Test of construct validity; 6) Establishment of final recommendations.

Results
Twenty-one morphometric measures were evaluated. Based on results of reliability and validity-testing including subjects with JIA (n = 70) and non-JIA controls (n = 19), seven measures received a "high recommendation" score. Those measures were associated with posterior mandibular height, occlusal cant, mandibular asymmetry, mandibular inclination, and anterior/posterior lower face height. Nine other measures were "moderately recommended" and five received a "somewhat recommendation" score. PLOS  Introduction domains: 1) description of dentofacial deformity, 2) treatment planning, and 3) long-term assessment.

Content-validity
Content-validity is defined as the level to which the content of the proposed measures reflects the underlying construct of the assessment. Six external experts, identified based on clinical expertise and research activity in this field, were invited to assess the content-validity of the measures defined in item generation. The external experts included three orthodontists (AK, CV, KDK) and three maxillofacial surgeons (SA, CMR, SEN). Content-validity testing was then performed by questionnaire, with assessment of each measure assessed by a Likert scale from 1-5 (1 = strongly disagree, 5 = strongly agree) in the three domains: 1) descriptive, 2) treatment planning, and 3) longitudinal assessment. An average score of 3.5 or above was considered a "high" score. For each measure, a unique statement was proposed for each domain: Descriptive domain. Statement: This is an important morphometric measure for the overall description of the severity of dentofacial growth deviation in JIA.
Treatment planning domain. Statement: This is a useful morphometric measure in the planning of orthodontic/surgical management.
Longitudinal assessment domain. Statement: This is an important measure for the description of changes over time for dentofacial growth deviation in JIA (e.g. pre-/posttreatment or with growth).
In addition, the external experts were asked to assess the overall "strength of recommendation" (SOR) for each measure using a 100 mm visual analogue scale (0 mm = not important, 100 mm = extremely important). Finally, the experts were asked to rank the measures based on "overall importance" from 1 to 21 (1 = most important, 21 = least important).
Subjects. To test content validity of the measures defined in item generation, the study group included patients with JIA. These patients represent the Aarhus TMJ arthritis cohort, which is a population-based group that contains standardized longitudinal observational data about patients with JIA. The included 70 patients are consecutive patients referred to the Section of Orthodontics, Aarhus University, Denmark for a CBCT from February 2011 to April 2014. Data was retrieved retrospectively. Inclusion criteria were: 1) a diagnosis of JIA according to ILAR criteria [27,28], and 2) a high quality full-face CBCT. Patients with a history of facial trauma or a congenital craniofacial anomaly were excluded. CBCTs obtained for routine orthodontic management from an age-matched non-JIA control group with no history of temporomandibular joint dysfunction were also included. Approvals from the Danish Health and Medicine authorities (3-3013-641/1) and the Danish Data Protection Agency (1-16-02-458-14) were obtained prior to the initiation of the study.
The retrospective retrieved CBCT examinations had been conducted in accordance with the manufacture instructions and in agreement with regulations approved by the Danish Health and Medicines authorities. A NewTom 5G, 18x16 cm field of view was used for all CBCTs. The image acquisition parameters included a scanning time of approximately 18 seconds, active radiation 3.6 sec (pulsed mode) with settings of 110 kV and 3-7 mA. All CBCT scans were constructed with a 0.30 mm isotropic voxel dimension. 3D data evaluation was conducted using Mimics software (Mimics1 18.0; Materialize Interactive Medical Image Control System, Leuven, Belgium). Anatomic landmarks were identified in all three planes of space (axial, sagittal, coronal). All proposed morphometric measurements were made for each subject by one author (CKI). Additionally, the radiographic appearance of each TMJ was categorized based on published criteria [17]: 1. Normal: Normal condylar shape with smooth and intact cortical outline/surface. 2. Abnormal: Condylar flattening or other changes in shape with smooth and intact outline and/or disruption of condylar outline with uneven surface due to cyst or erosion.

Test of reliability
To assess inter-rater variability, the same author repeated measurements for 30 subjects (selected using block randomization) a minimum of two weeks after the first assessment. These data were then used to calculate the smallest detectable difference for each measurement, defined as the smallest change that can be reliably observed between two consecutive observations (error of the measurement) [29,30]. To evaluate inter-rater agreement, a second author (PBS) evaluated the same 30 subjects.

Test of construct validity
Construct validity is defined as the degree to which the proposed outcome variables measure what they are intended to. To assess this, inter-group differences were calculated for each measurement. Associations between the inter-group results were then related to the following predefined hypotheses (H): H1: Morphometric measures demonstrate greater inter-side vertical and sagittal asymmetry in JIA subjects compared to controls.
H2: JIA 2 and JIA 3 subjects demonstrate greater inter-side vertical and sagittal asymmetry compared to JIA 1 subjects.
H3: Occlusal plane canting and occlusal plane inclination are more pronounced in JIA 2 and JIA 3 subjects compared to JIA 1 subjects and controls.
H4: Mandibular retrognathia is more pronounced in JIA 3 subjects compared to the other groups.

Establishment of final recommendations
A consensus-driven approach among all authors was used to establish final recommendations. Each measure was assigned a grade for each domain: highly recommended (+++), moderately recommended (++), somewhat recommended (+), not recommended (-). Because each domain was assessed separately, measures could receive higher recommendations for some domains than others. Statistical analysis. Descriptive statistics were computed. Intra-class correlations coefficients (ICC) were calculated using a two-way random effect model to assess intra-rater and inter-rater correlation. An ICC > 0.70 was considered acceptable. Bland-Altman plots and the limits of agreement were used to calculate smallest detectable differences [31]. Construct validity was tested against the predefined hypotheses using analysis of variance (ANOVA). For measures with a significant ANOVA result, independent Student's t-tests were then applied. The level of significance was set at p < 0.05. 23 landmarks, 12 internal reference planes, and six side-specific reference planes were defined ( Table 1, Table 2). From these, 21 morphometric measures were chosen for further evaluation: 1) Nine linear inter-side measurements (
H1 was accepted: larger inter-side differences in posterior mandibular height were observed in the three JIA groups compared to controls ( Table 5, No. 1 and 2). However, no significant inter-group differences in condylar height measures were seen (No.3).

Sella
The centre of the hypophyseal fossa S  H2 was accepted: greater inter-side vertical mandibular was seen in the JIA 2 and JIA 3 groups compared to JIA 1 subjects as illustrated by the inter-group differences in the mandibular axial angle (No. 12).

Midpoint between incisors Vertical midpoint between upper and lower incisors
H3 was partially accepted: Steeper occlusal plane canting (No. 8,9) was found in JIA groups compared to controls, but not between the 3 JIA groups. As hypothesized, occlusion inclinations were significantly steeper in the JIA 3 group compared to JIA 1 and controls (No. 16,17).  maximum score of "minor recommendation" within one or more domain (No. 4,5,7,10). The mandibular coronal angle was the only measure to receive "not recommended" in terms of therapeutic efficacy validity (No. 10). The highly recommended measures included: inter-side difference in posterior mandibular height, occlusal cant, mandibular asymmetry, mandibular inclination, and anterior/posterior lower facial heights. Table 7 depicts the seven highly recommended measures with a reference to the morphometric growth deviation each measure is intended to assess. The highly recommended measures represent seven unique aspects of abnormal dentofacial growth with no overlap (Table 7).

Discussion
Temporomandibular joint involvement is a frequent finding in JIA [2,3] and often impacts dentofacial growth. This belies the importance of establishing standardized recommendations for the evaluation of growth deviation in this population. In this study, we evaluated 21 morphometric measures using an established five-step method. To our knowledge, this is the first study to evaluate such measures in the JIA population. Seven measures received a high recommendation in all three domains and should therefore be considered of great importance for the study of dentofacial growth in JIA. These include: inter-side difference in posterior mandibular height, occlusal cant, mandibular asymmetry, mandibular inclination, and anterior/posterior lower facial heights. Future work will include establishing an index to assess the severity of dentofacial growth deviation in JIA for these seven highly recommended measures.
Despite many prior publications regarding 3D dentofacial imaging landmark identification and reproducibility, a systematic review from Pittayapat et al. concluded that additional standardization is necessary [9]. Prior studies fail to relate facial asymmetry with the error of the method. In the present study, we calculated the smallest detectable difference for each measure in order to define the minimum discernable change between two observations [29,30]. Surprisingly, we found high values for the smallest detectable differences of many previously published measures (Table 4). For example, 2D analyses concluded that condylar height was one of the most important measurements in the assessment of dentofacial growth deviation [25,32], but the smallest detectable difference for this measure in our study was found to be 5.7 mm. We hypothesize that this is due to large variation in landmark identification of the condylion point due either to condylar deformity from JIA or natural variability in condylar shape [33].
Construct validity was achieved by acceptance of the hypothesis that dentofacial growth deviation was more pronounced in the JIA groups compared to controls. However, contrary to our expectations, we did not observe inter-group differences between the three JIA groups in asymmetry-based measures like total posterior height 1 and 2 (No.1,2), and maxillary and mandibular occlusal canting (No. 8,9). This could be explained by the classification of the JIA groups based on radiographic appearance; contemporary theory explains the development of dentofacial growth deviation in JIA as a consequence of condylar growth disturbance rather than condylar damage [4,17,[34][35][36]. Therefore, the radiographic appearance of the condyle may be normal even when growth at that condyle has been impaired. This is in agreement Table 3. Content-validity. Six external experts rated the importance of each proposed morphometric measure (5-point Likert scale, endpoints: 1 = disagree, 5 = Strongly agree) within the three domians:1) description of dentofacial growth deviation (descriptive content), 2) Treatment planning, 3) Long-term changes validity. An average validity score ! 3.5 was considered of high content-validity. Additionally, the general importance is illustrated by the general strength of outcome measure (VAS 0-100 mm, endpoints: 0 = not important, 100 = extremely important) and rank of importance (1-21, 1 = most important, 21 = least important). with findings of Twilt et al., reporting dysmorphic mandibular development in patients with JIA without detectable condylar abnormalities on orthopantomograms [35]. It was also surprising that no differences were observed between the JIA groups and controls for condylar height (No.3), mandibular basal length (No.6), and Wits appraisal (No.20). This could also be due to the variation in landmark identification for these points, which is illustrated in the large values for smallest detectable differences for these variables. These findings reveal a pitfall of morphometric facial analysis which should be considered in future studies: statistical significance may not indicate clinical relevance if error of the method is not considered. This study has several limitations. Inter-side differences were used to express the degree of asymmetry [11,14,37], but were not related to normal variation associated with age. These results may therefore be misleading as, for example, a total posterior mandibular height interside difference of 4 mm may be a severe sign of asymmetry in an 8-year-old, but less significant in a 17-year-old patient. An alternative approach would have been to express asymmetry as a Table 4. Reliability tests. Intra-rater and inter-rater values based on 30 duplicate measurements. An intra-class correlation coefficient of ! 0.70 was considered acceptable. Abbreviations: ICC; intra-class correlation coefficients.

Smallest detectable difference
Inter-side diff. in bilateral linear distances ratio, but this would have limited clinical applicability. Additionally, we defined the "asymmetry-side", which serves as the basis for other calculations, as the side with the smallest total posterior mandibular height. However, this side may not have had the smallest linear distance or angle for all measurements. Nonetheless, this allowed consistent comparisons between sides and facilitated a complete characterization of facial asymmetry. Color-coded overlay mapping is an alternative technique for 3D visualization of facial asymmetry with promising clinical and research applicability [38,39]. As demonstrated by Liukkonen et al., mandibular asymmetry is common, even in a non-JIA population. This asymmetry is most often clinically irrelevant and frequently improves with age [40]. Although we demonstrated an inter-group difference in dentofacial symmetry Table 6. Final recommendations. Validity and reliability results were used for the establishment of final recommendation. Each morphometric measure was assigned with a grade of recommendation within each of the domains: highly recommended (+++), moderately recommended (++), somewhat recommended (+), not recommended (-). between JIA2/JIA3 and the control subjects, this exemplifies a general limitation to this study: there is a lack of standardized, consensus-based recommendations for radiographic examination in the JIA population. In our current practice, CBCT examinations are obtained when there is clinical suspicion for a dentofacial abnormality, such as a TMJ hard-tissue anomaly, a progressive dentofacial asymmetry, or for treatment decision making. Increasingly, CBCT is becoming routine in orthodontic and orthognathic surgical treatment planning and quality assessment. This expansion is fueled by the increasing availability of CBCT machines and the reduction in radiation dose required to obtain an adequate image, which is considerably lower than exposure from medical CT [41]. Recent publication by Markic et al. have compared alternative techniques for 3D visualization of facial asymmetry (orthopantomograms, CBCT, CT and magnetic resonance imaging (MRI)) and found equal abilities of the methods to assess inter-side difference in mandibular development [42]. Future research is needed to validate, if the morphometric measures recommended in the present study can be applied to MRI examination, allowing the elimination of ionizing radiation. and potentially the combination of soft and hard-tissue assessment into a single study.
In conclusion, we have identified and validated a series of morphometric measures for the assessment of dentofacial growth deviation in patients with JIA. Seven measures received a "high recommendation" score. Those measures were associated with posterior mandibular height, occlusal cant, mandibular asymmetry, mandibular inclination, and anterior/posterior lower facial height. These measurements will facilitate standardization of radiographic analysis in this population. This work offers important insight to the dentofacial consequences of TMJ arthritis in growing individuals and provides a framework for future research.