Evaluation of retinal and choroidal variations in thyroid-associated ophthalmopathy using optical coherence tomography angiography

Thyroid-associated ophthalmopathy (TAO) is a systemic autoimmune disorder or an organ-specific autoimmune inflammatory disease of orbital tissues. It is characterized by inflammatory cellular infiltration with lymphocytes, plasma cells, macrophages and mast cells. The most obvious pathological changes in orbital tissues include interstitial tissue edema, orbital fat hyperplasia and massively swollen extraocular muscles, which may cause orbital compression symptoms [1‐4]. Approximately 20% of patients with immune thyroid diseases will develop TAO, and 25–50% of TAO cases are closely related to hyperthyroidism, commonly termed as Graves’ ophthalmopathy, with higher morbidity in females [3, 4]. The exact pathogenesis of TAO remains unknown, but the clinical manifestations can be explained by the expansion of orbital volume due to autoimmune inflammatory infiltration. TAO can be classified into the following phases: an active phase with rapid progression and an inactive phase with symptom stabilization [5, 6]. However, approximately 3–5% of TAO patients in an inactive state will transit to an active state, which may result in aggravation of proptosis, lid retraction, dysfunctional eye motility, or even vision loss due to optic nerve compression [7, 8].

Different techniques have been employed to analyze the retinal and choroidal changes in TAO patients. Walasik-Szemplińska et al. [9] observed the alterations of ocular blood supply in TAO patients by color Doppler imaging. However, there are several influencing factors in the measuring process of ocular blood flow, such as eyeball movement and Doppler angles. The diameters of the examined vessels also limit its clinical applications. Optical coherence tomography (OCT) is a non-invasive imaging technique that used to visualize the detailed structures of the retina and choroid. Sayin et al. [10] found TAO patients had thinner inferior retinal nerve fiber layer (RNFL) thickness and macular thickness. Furthermore, choroidal thickness (CT) was found to be increased in TAO patients [11, 12]. With the advancement in technology, OCT angiography (OCTA) is a new promising imaging technique that is capable of imaging the retinal and choroidal vasculature noninvasively [13‐15]. Recently, OCTA has been widely used in ophthalmic diseases such as glaucoma and diabetic eye disease. However, few studies have investigated the retinal vessels in the macular region in TAO patients using OCTA.

A total of 82 eligible subjects were enrolled, including 20 active TAO patients, 33 inactive TAO patients, and 29 healthy participants. The basic characteristics of all participants are shown in Table 1. There were no significant differences in age, sex distributions, AL and CCT among the three groups (P = 0.339, P = 0.121, P = 0.100, P = 0.633, respectively). As expected, the proptosis in active TAO patients (21 ± 3) was highest, followed by inactive TAO patients (18 ± 3) and healthy controls (16 ± 2) (P < 0.001). Active TAO patients also had higher IOP than the other two groups (P < 0.001).

The traditional OCT analysis results are summarized in Table 2. The global average RNFL thickness was significantly different among the three groups (P < 0.001), post hoc pairwise comparisons revealed that active TAO patients had thinner RNFL thickness than the other two groups (P < 0.001, P < 0.001). Similar results were obtained when comparing the temporal and inferior RNFL thickness (P = 0.001, P = 0.001), while no significant differences were observed in superior and nasal RNFL thickness (P = 0.458, P = 0.117). With regard to CT in the macular region, TAO patients, no matter active or inactive, both had significantly higher CT than healthy individuals (all P < 0.05). However, no significant differences were detected between active and inactive TAO patients (all P > 0.05).

The evaluation of the superficial retinal vessels measured by OCTA are summarized in Table 3. The mean area of FAZ in the active TAO group was 0.36 ± 0.09 mm², which was significantly larger than the other two groups (P = 0.045, P = 0.001). In contrast, there was no significant difference between the inactive TAO group and control group (P = 0.130). However, the inactive TAO group had significantly higher vascular density than the other two groups (all P < 0.05), while there were no significant differences between the active TAO group and normal group (all P > 0.05). With regard to the perfusion density, significant differences were observed in the temporal and inferior areas (P = 0.045, P = 0.001), as well as the average values (P = 0.032). But the pairwise comparison results were not completely consistent.

To determine the potential influencing factors associated with these above parameters, Pearson’s correlation coefficients were calculated with the clinical variables of all the 82 eyes (Table 4). The FAZ area was positively correlated with IOP (r = 0.274, P = 0.013), while it was negatively correlated with axial length (r = − 0.344, P = 0.002). The vascular density and perfusion density were not significantly correlated with different clinical variables (all P > 0.05). Moreover, to determine whether these parameters can be used in TAO diagnosis, ROC curves were generated (Table 5). Table 6 showed sensitivities at fixed specificities and their cut-off values. The AUC analysis indicated that RNFL thickness had modest diagnostic power in active TAO/inactive TAO and active TAO/normal subgroups (AUC = 0.804, P < 0.001; AUC = 0.818, P < 0.001). Comparisons of CT yielded ROC curve areas of 0.814 for active TAO/normal and 0.828 for inactive TAO/normal (P < 0.001, P < 0.001). In contrast, FAZ area only exhibited a significant discriminatory power in active TAO/normal comparison (AUC = 0.711, P = 0.013). The vascular density and perfusion density also showed significant diagnostic ability in active TAO/inactive TAO and inactive TAO/normal subgroups (all P < 0.05), but exhibited a poor discriminatory power. Figures 2, 3 and 4 showed the detailed ROC curves for different subgroups.

OCTA is a new imaging technology that provides noninvasive fundus angiography, which works without contrast medium and avoids allergies and various contraindications. OCTA relies on the intrinsic motion of the fundus vasculature network to separate stationary structures to identify the blood flow. OCTA can also provide a 3D partition by comparing 2D images taken by indocyanine green angiography and fluorescein fundus angiography [18], which avoids artifacts and limitations such as a limited measurement time window and discomfort during inspection. OCTA is reliable, noninvasive, efficient, high quality and safe for fundus vascular imaging [18, 19]. Fundus perfusion depends on the orbital blood supply, and patients with TAO show orbital perfusion changes caused by pathological changes in orbital tissues. A previous study found certain hemodynamic changes in the ocular vasculature under Doppler imaging, and the condition in the ocular vasculature improved after orbital decompression [9]. Although color Doppler imaging has been widely used in vessel inspection, after certain ocular vasculature changes were detected, fundus perfusion changes can hardly be observed by Doppler imaging. Therefore, OCTA would be a perfect choice for further inspection to evaluate the status in fundus vessels.

In our present study, significant thinner RNFL thickness accompanied by higher IOP level was observed in active TAO patients, and the most affected RNFL quadrants were temporal and inferior quadrants. The superficial retinal vascular plexus may be the capillary network primarily responsible for RNFL nourishment. In our study, the macular superficial vessel density remained unchanged in active TAO patients while the thinning of RNFL was observed, suggesting that losses in the peripapillary RNFL in active TAO patients might have been influenced by factors other than retinal microcirculation. Active TAO patients often suffer from secondary compressive IOP rise, thus leading to RNFL defects [20, 21]. The enlargement of extraocular muscles and increase of orbital soft tissue volume could cause direct compression of the optic nerve, which also contribute to dysthyroid optic neuropathy. Earlier detection of RNFL thinning would suggest the presence of optic neuropathy, indicating its use in the evaluation of this disease profile.

Significant greater CT was observed in active and inactive TAO eyes as compared to the normal eyes. Çalışkan et al. [6] found the subfoveal CT in active TAO patients was significantly greater than those with inactive TAO or healthy individuals, even after adjusting for age, axial length and IOP. Similar results were observed in another study conducted by Özkan et al. [22]. Besides, Yu et al. [11] also identified increased CT in TAO patients at different locations in the macular region. Theoretically, one possible explanation for the choroidal variations might be the venous obstruction and congestion, caused by reduce orbital venous drainage, which was the result of increased retrobulbar pressure [22, 23]. Furthermore, the hyperdynamic cardiovascular state of hyperthyroidism could induce an increase in cardiac output, possibly affecting the choroidal perfusion [24].

OCTA has been widely used to analyze the detailed characterization of the retinal and choroidal vasculature in the macular and peripapillary regions [25‐27]. Due to the limitations of the analysis software, only superficial vascular plexus in the macular region could be quantitatively analyzed in our study. The FAZ area is a capillary-free area in the central macula that serves as the most sensitive part of the retina. In our study, the FAZ area was significantly enlarged in active TAO patients. But we didn’t detect disintegrity of the vascular arcades surrounding the FAZ area. Previous studies reported that the enlargement of FAZ area more objectively supported the findings of capillary nonperfusion [28, 29]. However, our findings suggested the superficial vascular plexus was increased in inactive TAO patients. The FAZ area measurements showed high reproducibility and repeatability, but the real data vary in different studies. Several possible confounding factors, such as age, sex, spherical equivalent, and axial length, may influence the size of FAZ area [30, 31]. In our study, the FAZ area was positively correlated with IOP (r = 0.274, P = 0.013), while it was negatively correlated with axial length (r = − 0.344, P = 0.002), which might give a partial explanation.

The macular vascular density and perfusion density of the superficial layer were quantitatively evaluated in TAO patients. Inactive TAO patients had significantly higher vascular density than that in active TAO and controls. With regard to perfusion density, the pairwise comparison results were not completely consistent. Collectively, the data analysis revealed that inactive TAO patients seemed to have greater perfusion density. In previous studies, Ye at al [32]. reported that active TAO patients presented with an increased retinal microvascular density. Akpolat et al. [33] demonstrated that the temporal and nasal parafoveal vessel density was significantly higher in inactive TAO group, which was consistent with our results. However, Tehrani et al. [34] found active and inactive TAO patients both had significantly lower superficial vessel density in the nasal parafoveal sector. Mihailovic et al. [35] also showed inactive TAO patients had decreased vessel density in the superficial OCT angiogram. In these studies, the normal controls were not totally matched to the study population on several nonexperimental factors, such as age, sex, and axial length, which might at least partially explain the discrepancy. Besides, technical differences and variations in patient status, such as inactive or active TAO state or moderate-to-severe status, may also contribute to the incongruity. Further studies with more rigorous design are desiderated to explore the alterations in TAO patients.

These retinal and choroidal changes might be correlated with variations in orbital blood flow. Doppler imaging of orbital vessels revealed that the resistance index (RI) in the ophthalmic artery (OA) was decreased inactive TAO patients, but systolic velocity remained unchanged, suggesting increased blood flow in OA in inactive TAO patients. The RI in the central retinal artery (CRA) was increased, and the velocity and RI in superior ophthalmic vein (SOV) showed no difference compared with that in the control group [9]. The increased blood supply in the ocular vasculature may partly explain the increased vessel density. Walasik-Szemplińska D et al. [9] found that RI in OA decreased in active TAO. Velocity and RI were increased in CRA, while increased RI and decreased velocity were detected in SOV, indicating circulatory disorder in the ocular vasculature. Reverse flow was also observed in SOV, indicating severe stasis in SOV, which usually correlated with enlarged extraocular muscles. The SOV was considered to play important roles in the inflammatory stage in TAO, and recent studies have demonstrated blood flow reduction in the SOV during active TAO, indicating orbital circulation disorder was resulted from a total effect of increasing venous pressure and high RI, which was caused by elevated intraorbital pressure. Autoimmune inflammation in orbital tissues, including interstitial tissues, orbital fat and extraocular muscles, was the main cause of the high intraorbital pressure [3]. Moreover, Onaran et al. [36] observed a reduction in SOV flow among patients after orbital decompression along with the disappearance of the reverse flow. Therefore, we proposed that vascular physiological changes and elevated intraorbital pressure caused by the direct effect of autoimmune inflammation on ocular vessel and orbital tissues lead to variations in fundus blood flow in active TAO, along with effects on RNFL thickness, CT, FAZ, vessel density and perfusion density, as observed in our study.

In the ROC analysis, OCT-derived RNFL thickness and choroidal thickness showed apparent diagnostic ability in TAO, which were consistent with previous studies [10, 18]. The FAZ area, vascular density and perfusion density also exhibited a significant discriminatory power to distinguish between TAO patients and controls. Ye at al [32]. also showed high diagnostic power to differentiate active TAO patients from normal controls, with AUC of 0.97 for superficial retinal density and AUC of 0.8 for deep retinal density. We hypothesize that these parameters may have the predictive value in the diagnosis of TAO. In addition, non-invasive measurements of these parameters are easily accepted by the patients. Clearly, these parameters were poor markers, further investigations are needed to substantiate these findings in a much larger cohort.

As a preliminary study, our present findings have several limitations. First, this was a cross-sectional study without follow-up data, which prevented us to correlate the vascular changes with the disease progression. Second, the morphology of the superficial retinal vessels is not equal to the hemodynamic changes, which may limit our understanding of the pathogenesis. Third, analysis of the retinal vessel parameters was limited to the superficial layer due to the limitation of analysis software, further investigations of deep retinal vessels should be performed to better demonstrating the retinal vessel variations. Fourth, the sample size was relatively small due to the rigorous selection standards. Moreover, the study group included in the current analysis were TAO patients. Dysthyroid patients in the absence of TAO were not enrolled, which may also limit the interpretation of the results to some extent. Nevertheless, our results indicated retinal and choroidal variations in TAO patients, further researches are highlighted to supplement and extend these preliminary results.

TAO patients had significant variations in RNFL thickness, choroidal thickness, FAZ area and superficial retinal vessels. These parameters appeared to be potential adjuncts in the evaluation of TAO patients.