Purpose: Identify a reproducible measure of axial globe position (AGP) for multicenter studies on patients with thyroid eye disease (TED).
Methods: This is a prospective, international, multicenter, observational study in which 3 types of AGP evaluation were examined: radiologic, clinical, and photographic. In this study, CT was the modality to which all other methods were compared. CT AGP was measured from an orthogonal line between the anterior lateral orbital rims to the cornea. All CT measurements were made at a single institution by 3 individual clinicians. Clinical evaluation was performed with exophthalmometry. Three clinicians from each clinical site assessed AGP with 3 different exophthalmometers and horizontal palpebral width using a ruler. Each physician made 3 separate measurements with each type of exophthalmometer not in succession. All photographic measurements were made at a single institution. AGP was measured from lateral photographs in which a standard marker was placed at the anterior lateral orbital rim. Horizontal and vertical palpebral fissure were measured from frontal photographs. Three trained readers measured 3 separate times not in succession. Exophthalmometry and photography method validity was assessed by agreement with CT (mean differences calculation, intraclass correlation coefficients [ICCs], Bland-Altman figures). Correlation between palpebral fissure and CT AGP was assessed with Pearson correlation. Intraclinician and interclinician reliability was evaluated using ICCs.
Results: Sixty-eight patients from 7 centers participated. CT mean AGP was 21.37 mm (15.96-28.90 mm) right and 21.22 mm (15.87-28.70 mm) left (ICC 0.996 and 0.995). Exophthalmometry AGP fell between 18 mm and 25 mm. Intraclinician agreement across exophthalmometers was ideal (ICC 0.948-0.983). Agreement between clinicians was greater than 0.85 for all upright exophthalmometry measurements. Photographic mean AGP was 20.47 mm (10.92-30.88 mm) right and 20.30 mm (8.61-28.72 mm) left. Intrareader and interreader agreement was ideal (ICC 0.991-0.989). All exophthalmometers' mean differences from CT ranged between -0.06 mm (±1.36 mm) and 0.54 mm (±1.61 mm); 95% confidence interval fell within 1 mm. Magnitude of AGP did not affect exophthalmometry validity. Oculus best estimated CT AGP but differences from other exophthalmometers were not clinically meaningful in upright measurements. Photographic AGP (right ICC = 0.575, left ICC = 0.355) and palpebral fissure do not agree with CT.
Conclusions: Upright clinical exophthalmometry accurately estimates CT AGP in TED. AGP measurement was reliably reproduced by the same clinician and between clinicians at multiple institutions using the protocol in this study. These findings allow reliable measurement of AGP that will be of considerable value in future outcome studies.