This study performed a retrospective analysis of serial optic disc photographs obtained from subjects aged younger than 18 years who were consecutively enrolled from a database of patients first examined for suspected glaucomatous optic disc between October 2004 and October 2014 at Seoul National University Bundang Hospital. This retrospective, observational study was approved by the institutional review board of Seoul National University Bundang Hospital (B-1705/396-110). The requirement for informed consent was waived owing to the retrospective nature of the study. The study followed the tenets of the Declaration of Helsinki.

Each subject underwent ophthalmic examination that included best-corrected visual acuity, cycloplegic refraction, slit-lamp biomicroscopy, and disc photography. Subjects were eligible for inclusion when their medical records included serial color- or red-free fundus photographs with an interval between the first and last photographs of at least 1 year. Eyes with a history of intraocular surgery, a history or evidence of glaucoma, other optic neuropathies or retinal disease, an elevated intraocular pressure (IOP), or a spherical equivalent (SE) of less than −6 diopters were excluded. The medical records of selected subjects were reviewed, and data on sex, age, laterality, and cycloplegic refractive error at the initial and final examinations were collected. When both eyes were eligible for inclusion, one eye was randomly selected.

This study initially included the 398 eyes of 199 subjects who had more than 1 year of follow-up using serial disc photography. Of these, 65 eyes were excluded because of a history or evidence of congenital glaucoma (n = 16), juvenile glaucoma (n = 21), or other optic neuropathy (n = 10), or SE <−6 diopters (n = 18), leaving a final sample of 333 eyes of 173 subjects. When both eyes were eligible for inclusion, one eye was randomly selected.

Subject age and refractive error at the time of initial fundus examination were 6.8 ± 1.7 years (range, 2 to 11 years) and 0.2 ± 2.6 diopters (range, −6.0 to +5.5 diopters), respectively. The follow-up period was 44.8 ± 21.1 months (range, 12 to 103 months). The coefficient of the intraobserver repeatability for measuring AVDA was 4.5°. Based on this value, progressive torsion was defined as an AVDA change exceeding 5°.

Of the 173 finally included eyes, 42 (24.3%) had progressive torsion and were classified as the torsion group; the remaining 131 eyes were classified as the non-torsion group. Of the 42 eyes in the torsion group, 15 and 27 had superior and inferior torsion, respectively, as defined by clockwise and counterclockwise rotations, respectively, of the vertical diameter axis at a right-eye orientation (Fig. 3). Considerable torsion was found in 21 eyes (12.1%).

Table 1 compares the baseline characteristics between the torsion and non-torsion groups. No significant difference was found between groups with regard to age, sex, laterality, or follow-up period. The final SE was more myopic in the torsion group than in the non-torsio n group (p = 0.004), and the degree of myopic shift was also significantly greater in the torsion group than in the non-torsion group (−2.4 ± 1.7 vs. −1.4 ± 1.6 diopters, p = 0.001).

Table 2 compares changes in the optic disc and parapapillary region between the two groups. The final HVDR was smaller in the torsion group than in the non-torsion group (p = 0.005). In addition, the change in HVDR was significantly larger in the torsion group (p < 0.001). PVDR did not differ significantly between the two groups on initial examination (p = 0.183), but the subsequent change in PVDR was significantly larger in the torsion group (p < 0.001).

The torsion in 19 of the 21 eyes with considerable disc torsion was fully explained by the tilt of the optic disc on an oblique axis. In the remaining eyes, torsion could not be explained solely by tilt, and additional rotation was considered to be present in those eyes (Fig. 2D–2F).

In univariate analysis, torsion was significantly associated with the last measured value of SE (p = 0.004), changes in SEs (myopic shift, p = 0.001), and changes in HVDR and PVDR during follow-up (p < 0.001 and p < 0.001, respectively). Since changes in HVDR and PVDR represent different aspects of optic disc changes associated with myopic shift [20], multivariate analysis was performed in two ways to avoid multicollinearity between them. Changes in HVDR and PVDR during follow-up remained statistically significant in multivariate analysis (p < 0.001) (Table 3).

We identified 42 cases in which progressive optic disc torsion occurred during the study period. Torsion was associated with disc tilt and the development/enlargement of PPA in all eyes. Additional rotation of the optic disc axis was observed in two patients. To the best of our knowledge, this study is the first to document the development of disc torsion using serial fundus photographs.

We previously demonstrated the presence of optic disc tilt in children with incipient myopia [20]. Tilt was accompanied by the development/enlargement of PPA and was associated with myopic shift, suggesting that disc tilt derives from scleral stretching secondary to axial elongation. In the present study, torsion was universally associated with a decrease in HVDR (i.e., disc tilt), and also with the development/enlargement of PPA and myopic shift. These findings suggest that torsion represents tilt on an oblique axis. Axial elongation can occur in various directions [21], and the direction of axial elongation can influence the direction of disc tilt. For instance, when the posterior center of the axis of elongation is located superotemporally, scleral stretching would occur in the superotemporal direction, resulting in PPA in the superotemporal region, with oblique disc tilt being perpendicular to the direction of scleral stretching. The obliquely tilted disc would have an oblique disc axis, and be identified as torsion.

The changes in almost all cases in the torsion group were explainable by oblique disc tilt. However, in two eyes torsion could not be explained by oblique tilt because the nasal disc margin was deviated from the expected disc margin after oblique disc tilt. This finding suggests that true rotation of the disc axis also occurs in some eyes. Although we do not have a clear explanation for such rotation, we speculate that it is attributable to axial elongation in multiple directions. Such axial elongation could result in complex changes in the optic disc, resulting in apparent rotation of the optic disc axis.

Torsion is a common feature in myopic eyes [15], and it has recently become a focus of interest in relation to how it influences the development of GON. In young patients with unilateral myopic glaucoma, greater torsion was found in the affected eye. In particular, it is intriguing that the direction of torsion was related to the location of retinal nerve fiber layer (RNFL) defect: this defect was seen in the superotemporal sector in eyes with superior torsion, and in the inferotemporal sector in eyes with inferior torsion [15]. Our findings may provide a plausible explanation for the relationship between optic disc torsion and the location of glaucomatous damage. The superotemporal and inferotemporal sectors are preferentially involved in glaucoma due to the presence of less support for the axons from connective tissue [22]. In eyes with optic disc tilt, tensile and/or shearing stresses arising from scleral stretching may induce axonal damage. In this process the direction of scleral stretching may result in stress varying by region of the optic nerve head. For instance, in eyes with superior torsion, the change would be largest near the superior disc margin and thereby increase the possibility of axonal damage in the superior optic nerve (and vice versa for eyes with inferior torsion).

An RNFL defect was not detected on follow-up photographs in any of the eyes in the torsion group in this study (data not presented). However, especially in myopic patients aged in their 20s or early 30s, RNFL defect is not rarely seen despite low IOP, particularly among those with disc torsion [2324]. We speculate that the tensile stress derived from torsion does not immediately lead to development of axonal damage. However, eyes with torsion will be subject to sustained tensile stress, which may eventually damage the axons. An analogy to this finding may be found in the relationship between ocular hypertension and glaucoma. Eyes with high IOP do not necessarily exhibit optic nerve damage (i.e., ocular hypertension) at initial presentation, but some develop glaucomatous damage years later [25].

AVDA was larger than 15° at baseline in 10 of the 173 included eyes. These eyes all had PPA and a tilted disc appearance. We speculate that those discs were torted before the baseline photograph was obtained. Progressive considerable torsion was not observed during the study period in these 10 eyes, and so the incidence of considerable disc torsion may have been underestimated in this study.

The relationship between myopia and glaucoma remains unclear. It is generally acknowledged that myopia is a risk factor for glaucoma. In contrast, recent studies have demonstrated that glaucoma progression does not occur more rapidly in myopic eyes [2627]. Our data may provide some insight into this puzzle. The tensile stress associated with torsion may increase the susceptibility of axons to glaucomatous damage. Thus, it is possible that eyes subject to torsion will develop glaucoma despite the relatively low IOP-induced stress not being sufficient to induce glaucomatous damage in eyes without torsion. Meanwhile, the tensile stress would remain focused near the superior or inferior pole of the optic disc, depending on the direction of torsion. Thus, once glaucomatous damage occurs in the respective polar area in these eyes, the remaining axons may remain healthy if the glaucomatous insult (i.e., IOP-induced stress) is not sufficient to induce glaucoma by itself. In this situation the rate of disease progression would decrease. This may be a factor contributing to the slow rate of disease progression in myopic patients.

This study was subject to several limitations. First, we did not measure axial length, and myopic shift was used as a surrogate of axial elongation. However, myopic shift is attributed to axial elongation that occurs in childhood [28]. Second, we could not measure the true length or distance values in fundus photographs because we could not correct for the magnification effect due to the actual axial length being unknown. However, all measurements in the study were of angles and ratio parameters that would not be affected by magnification errors. Third, the subjects were glaucoma suspect patients, and so the findings of this study might not be applicable to a general population. Fourth, the subjects were not observed until the end of eye growth. It is possible that the non-torsion group could show changes in the future. Lastly, disc changes were evaluated based on two-dimensional analyses. This was because three-dimensional measurements of the optic nerve head were not feasible in most of the children, who were aged 6.8 ± 1.7 years at the time of baseline examination. Three-dimensional analysis may provide greater insight into optic disc tilt. Nonetheless, our data clearly indicate that optic disc torsion can be an acquired feature arising from scleral stretching.

In conclusion, we have demonstrated the development of optic disc torsion in children. Torsion was mostly explained by optic disc tilt on the oblique axis. The findings provide a framework for understanding torsion-related GON.