Evaluation of Globe Position Asymmetry in Endocrine Orbitopathy Patients and a Control Group – A Computed Tomography–Based 3D Cephalometric Analysis

Endocrine orbitopathy (EO) is a disfiguring, potentially vision-threatening disease affecting mostly women from 30 to 60 years of age.^1,2 It is a typical extrathyroidal symptom of Graves’ disease and may cause exophthalmos, eyelid retraction, chemosis, periorbital tissue increase, diplopia, and conjunctival injection.^1,2 Recent findings suggest that asymmetry of globe position is a predictor of disease progression and severity.^3,4 By now, assessment of globe position (and thus asymmetry) relative to the orbit is performed mainly by way of exophthalmometry^5-7 or, to a lesser extent, by using 3D imaging.^8-10 Besides providing the clinician with knowledge on the possible presence of asymmetry, this assessment might be an important aspect in treatment planning if decompression surgery is scheduled to achieve a symmetric esthetically pleasing outcome.

As the authors’ workgroup had developed a 3D cephalometric computed tomography (CT)-based orbital analysis,¹¹ they sought to evaluate the prevailing asymmetry in a database of patients with and without EO, measure the clinical usability of the 3D analysis, and compare the data to previous reports.

The 3D cephalometry was based on FAT software^12,13 that uses the open-source VTK (Visualization Toolkit) library (vtk.org). The interface has been structured to permit best clinical usability. After loading and registering the CT data, a set of anatomical landmarks (Table 1) was placed either directly on the 3D rendering or by using crosshairs on the 3D image or 2D sagittal, coronal, and axial slices. The software could zoom the images, determine threshold indices, and rotate the images automatically for each landmark and highlight the supposed area were the landmark would be placed by using the inherent software landmark library. After all landmarks were placed and their positions confirmed, the software would run a cephalometric script to create lines and planes or construct new landmarks. The results were then displayed, consisting of distances, angles, proportions, and symmetry measurements in a spreadsheet ready for export. Screenshots could be captured if appropriate.

For measurement of globe position symmetry, the recorded distances are stated in Table 1. For the analysis, 36 landmarks were used: 14 bilateral and 1 midline landmark, and 2 constructed bilateral and 3 constructed midline landmarks (Table 1, Figure 1). A total of 12 planes were constructed and 33 distances were measured (16 bilateral, 1 midline) (Table 1). As the software was modular, existing landmark sets could be reused, landmarks added, and the cephalometric analysis modified.

For the investigation, CT images obtained from 52 patients with bilateral EO were used. Mean age of these patients was 49.1 ± 10.4 years (range, 28-76 years); 35 patients were female and 17 were male. For the control (non-EO) group, CT images were obtained from 54 patients without known orbital/craniofacial pathology. Mean age of these patients was 41.6 ± 18.3 years (range, 18-84 years); 21 patients were female and 33 were male. CT data did not exceed 1.5 mm thickness of contiguous slices.

Statistical analysis was performed using WinMEDAS software (Fa. Christian Grund). Symmetry was assessed with paired t tests, and comparisons regarding sex or EO were performed with unpaired t tests. Correlations were measured using Spearman’s rank correlation test. Statistical significance threshold was set at P ≤ .05.

Regarding average time needed to perform a single analysis from a CT scan, approximately 5 minutes were spent on data selection, loading, analysis, and export. As shown in the descriptive statistical results (Table 2), patients in the EO group presented a statistically significant larger globe protrusion regarding the bony orbit (eg, orbital apex), orbital aperture, and soft tissue lateral rim (computed Hertel values). 

Asymmetry was evaluated in 3 ways. First, the mean values of both sides were compared. Here, the non-EO group presented minor, statistically nonsignificant differences (Table 2). In the EO group, both sagittal side differences and the lateral globe position in relation to the orbital aperture midpoint showed statistically significant asymmetry values. Next, individual difference values were computed for each patient. Here, the EO group showed statistically significant larger values, too. For the bony sagittal value, asymmetry of globe protrusion were between 2 and 2.2 mm, depending on the measured distances; in contrast, these values were between 0.7 and 1.2 mm in the non-EO group. Regarding interglobe distance and globe distance to the midsagittal plane, absolute values and asymmetry were also statistically larger in the EO group. Only in the cranial/caudal deviation to the Frankfurt horizontal plane were no statistical differences for absolute and asymmetry values found. Lastly, the individual asymmetry values in the sagittal dimension (exophthalmos) were divided in 3 groups: asymmetry <2 mm, asymmetry from 2 to 4 mm, and asymmetry >4 mm. Here, the EO group showed larger percentages of patients in the 2- to 4-mm group and the >4-mm group. In the control group, no patients had asymmetry values >2 mm (Table 2). Due to the rather small vertical and lateral asymmetry values, no subgroups to define the extent of asymmetry were calculated.

In both the EO and non-EO groups, the right side was more often the prominent one (EO group: 30 patients with right predominance vs 20 patients with left predominance; non-EO group: 32 CTs showed right-side prominence vs 22 with left-side prominence).

Concerning asymmetry prevalence and sex, no statistically significant correlation was found in the non-EO group. In the EO group, male sex was correlated with asymmetry in the distances globe to orbital apex, globe lateral position, globe position anterior, globe protrusion, and bony orbit to the frontomalar suture (FMS) as calculated by Hertel exophthalmometer. (Table 3).

Regarding correlations between the different globe position measurements, the following results were seen: sagittal globe position measured in relation to the orbital apex (directly measured or projected) correlated statistically significantly to the sagittal globe position with respect to the orbital aperture; higher correlation coefficients were seen in the EO group. Interglobe distance showed no or a weak significant correlation to the sagittal globe position measured in the orbital aperture region. Sagittal globe position to the orbital apex plane presented no statistically significant correlation with the lateral and horizontal globe position.

Summarizing the findings, patients in the EO group presented a statistically significantly larger globe protrusion and a higher degree in asymmetry, both in average and when grouped according to asymmetry severity. Male sex correlated with some asymmetry parameters, and the right side was more often the prominent eye. Sagittal globe position with respect to the orbital apex correlated with sagittal globe position calculated to the orbital aperture.

Sagittal globe position in healthy persons varies depending on age, sex, and ethnicity.⁵ For clinical measurements, various devices have been invented but the most widely used are Hertel and Naugle types; several modifications, however, have been suggested (eg, the Luedde or Mourits meter).^14,15 When using Hertel devices for these measurements, parallax errors, rotation on the horizontal plane, and incorrect resting of the footplate on the lateral orbital rim¹⁵ may lead to errors in the range of ±1 mm.¹⁴ Other devices may furnish differing results in singular examinations, but these devices seem to be comparable when used for large groups of patients⁷ and have been shown to provide acceptable interobserver reliability.⁶  According to previous reports,^3-5,16 asymmetry in >90% of individuals in healthy study groups is <1 mm and never >2 mm.

CT has been employed to measure sagittal globe position in various investigations.^8-10 In all studies, CT results showed a high correlation with clinical Hertel values. A noteworthy difference between these investigations and the current study, however, lies in the CT analysis. For sagittal globe position calculation in previous investigations, the patient had to be oriented parallel to the Frankfort plane by the radiologist, and the acquired axial planes were used to calculate proptosis values. Perfect orientation may be performed in selected studies but may be problematic in routine investigations, especially if CTs are performed at a different facility than where measurements would be performed. In the current investigation, patient head orientation was definitively mixed. Thus, using 3D cephalometry and image manipulation, a sound analysis was possible; this would not have been the case if the original axial CT planes had been utilized. It has been shown that landmark repeatability is high in anatomically defined landmarks^17,18 (ie, Bookstein types 1 and 2^19,20), leading to a low inter- and intraobserver variability of 0.5 to 1 mm.^21-25  Thus, linear measurements should be precise and reliable.²⁶

Comparing the authors’ asymmetry measurements with clinical exophthalmometric results, the mean values for both sides showed differences in accordance with previous reports.^5,16 These variations were <1 mm in both groups with higher values seen in the EO group (0.9 and 0.8 mm in EO patients; 0.2 and 0.1 mm in the non-EO group; Table 2). Sagittal side differences were statistically higher in the EO group.

Looking at asymmetry values on an individual basis, the situation was different. Here, patients in the EO group presented asymmetry of >2 mm in 38% and 42%, respectively, for the distances from anterior globe to apex orbitae and anterior globe to cor-apex orb plane (Table 3). This is in accordance with several investigations^3,4,27 that stated that asymmetry is not uncommon in patients with EO. Panagiotou et al³ reported that asymmetry >2 mm may be encountered in 9% to 34% of patients with bilateral EO disease, whereas Kavoussi et al²⁷ found asymmetry in 38%. This corresponds with the current study’s finding of asymmetry of >2 mm in 38% and 42% of patients, depending on the calculation performed (Table 3). The results of the EUGOGO study,⁴ which reported asymmetry in 30.9% of patients with EO, are difficult to compare because a score of 6 symptoms was used to define asymmetry, one of which was a proptosis difference of >2 mm. As with the present study, Perros et al and Kavoussi et al^4,27 found male sex  to correlate with pronounced asymmetry; the reason for this correlation is only speculative. The same also applies to the dominance of the right side in asymmetry cases for which no causes are known.

Novel findings of the present study were the amount of severe asymmetry (>4 mm side difference) and the slight but statistically significant asymmetry regarding the lateral globe position. Furthermore, the authors could demonstrate the correlation of sagittal globe position measured to the bony orbit with the sagittal position in relation to the bony aperture, whereas sagittal position showed no correlation with the lateral, horizontal, or vertical globe position. Concerning the higher correlation coefficients of proptosis measured to the deep bony orbit and the orbital aperture in the EO group, a possible explanation could lie in the variable lateral inclination of the orbital aperture.¹¹ In individuals in the non-EO group, this anatomical trait was an influencing factor of sagittal globe position measured to the orbital aperture, thus leading to a comparatively low correlation coefficient. In patients with EO who had pronounced proptosis, however, the influence of the orbital aperture tilting (eg, retrusion of the lateral orbital rim) could become less important in the statistical correlation analysis.

In the non-EO group, no asymmetry >2 mm was found, which is in accordance with  previous studies.^3,5,16

Even if this 3D cephalometric approach worked well, there are 2 potential problems in performing 3D cephalometric globe position analysis. First, reference planes were used for measurements. These are necessary if true sagittal, vertical, or horizontal values are to be recorded. These planes rely on symmetrical landmarks; therefore, stable, symmetric, and clearly identifiable landmarks have to be selected. In orthodontics and maxillofacial surgery, bilateral points (such as Porion and FMS), midline landmarks (such as Nasion and Sella), or midpoints of symmetrical landmarks are used.^18,24 Minimal asymmetry in selecting symmetrical landmarks can thus lead to asymmetric values measured to these planes. Performing measurements in larger study groups should eliminate this minor tilting of the reference planes, but in singular measurements, this could have an impact in asymmetry scores. The second and probably more relevant problem lies in identifying anterior and posterior globe landmarks. These landmarks were the most difficult to define within the current analysis. According to the Bookstein definition, these are type 3 landmarks and therefore, by nature, are less precise.^19,20 The authors tried to eliminate this factor by using crosshairs and simultaneous scrolling in three 2D windows in addition to reformatting or tilting the axial planes if needed. However, alternative ways to measure anterior and posterior globe landmarks could be required to reduce the effort of landmark placing (which should be at least semiautomatic to ease usability and reduce investigation effort). By now, no such algorithms are known to the authors.

Summarizing the aforementioned effects, future software should incorporate automatic orientation plane extraction by determination of multiple symmetrical landmarks and an algorithm to define the “true” anterior and posterior globe landmarks (eg, measured to the true vertical plane as defined in orthodontic analyses).

To the authors’ knowledge, there is no study utilizing 3D analysis of the orbit and globe position that has presented information on the time required for a single analysis. It can be speculated that this ranged well above the average 5 minutes required by the authors’ software. Nevertheless, clinical exophthalmometry using a Hertel device is faster and less expensive and does not require 3D data. Thus, in a first investigation on proptosis and asymmetry, clinical devices should be the method of choice. In case of pathologic findings that will lead to 3D image acquisition, however, a fast and reliable 3D analysis will deliver precise information on the bony orbit and soft tissues¹¹ on globe position and asymmetry.

Importantly, 3D cephalometry could be used to analyze treatment effects on globe position. The tool would be well-suited to investigating treatment changes beyond a postsurgical reduction of clinical Hertel values because surgery might influence the correct position of the exophthalmometer.

In conclusion, previous findings reporting on higher occurrence of sagittal globe position asymmetry were confirmed, as were the higher prevalence of asymmetry in male patients and the predominance of right-sided asymmetry. By way of 3D cephalometry, novel information on the concurrent lateral globe displacement was found, demonstrating that EO causes a complex change in globe position. Furthermore, the extent of asymmetry (13% of patients in the EO group had asymmetry >4 mm) has not been previously reported in the literature. As asymmetry has been reported to correlate with disease progression or severity,^3,4 a 3D analysis based on CT data might be appropriate in these cases. When planning treatment with decompression surgery in patients with EO, clinicians must consider the impact of such asymmetry, which could be quite prominent. Here, software-supported 3D cephalometry can be an efficient means to analyze orbital hard and soft tissue anatomy in patients with EO  and could be an important part of any surgical software, even if some features (like the exact definition of the most anterior globe landmark) need  improvement. Such future refinements will be the task of the authors’ workgroup. Furthermore, having a precise tool for globe position measurement could be beneficial when evaluating the progress of presurgical asymmetry after decompression.

Affiliations: ¹Department of Ophthalmology, Leipzig University, Leipzig, Germany; ²Department of Oral & Maxillofacial Plastic Surgery, Helios Vogtland-Klinikum Plauen, Plauen, Germany; ³Department of Informatics and Media, Leipzig University of Applied Sciences, Leipzig, Germany; ⁴Department of Oral & Maxillofacial Plastic Surgery, Leipzig University, Leipzig, Germany

Correspondence: Thomas Hierl, MD, DDS, PhD; thomas.hierl@helios-gesundheit.de

Sources of support: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The development of the 3D analysis software (FAT) was partially funded by German Federal Ministry of Economics and Technology ZIM KF2036708SS0.

Ethics:  Ethical approval for this study was given by the Leipzig University ethics commission (No. 285-14-25082014), and the study was performed in accordance with the Declaration of Helsinki on medical protocols and ethics. As pseudonymization of the CT data was performed, no requirement of informed consent was requested by the ethics committee.

Disclosures: The authors declared no conflicts of interest with respect to the research, authorship, and/or publication of this article.