Evaluation of Vitreoretinal Cellular Inltration in Uveitis on Optical Coherence Tomography

This study aimed to retrospectively evaluate vitreoretinal cellular inltration in uveitis on spectral domain optical coherence tomography (SD-OCT). Forty eyes of 26 patients with uveitis in the posterior segment and 22 eyes with no apparent abnormalities from 22 individual controls were enrolled. SD-OCT cross-sectional images through the fovea in all subjects were graded as follows: grade 0, absence of cell-like hyperreective particle (CLHP); grade 1, ≤ 10 CLHP on the retinal surface without vitreous CLHPs; grade 2, retinal CLHPs with ≤ 5–10 CLHPs in the vitreous; grade 3, 11–50 vitreous CLHPs; and grade 4, ≥ 50 vitreous CLHPs with diffuse distribution. The sensitivity and specicity for detecting active intraocular inammation in the classication with the presence or absence of the positive OCT-based grading scores (grade ≥ 1) were 85% and 95.5% respectively. Moreover, the OCT-based grading score was signicantly correlated with the vitreous haze classication (R = 0.35, p = 0.029). The interobserver exact agreement of the OCT-based grading score was moderate (κ = 0.50). Thus, this newly developed OCT-based grading system for vitreoretinal cellular inltration could a potential biomarker to evaluate inammation in the posterior segment of the eye.


Introduction
Uveitis can be caused by more than 100 disorders associated with intraocular in ammation [1]. Some have infectious etiologies, and another are caused by systemic disease, including autoimmune diseases, although the causes can be unknown or unidenti ed in many cases. Uveitis is also well known worldwide as a disease that potentially leads to blindness and accounts for 25% of cases of total social blindness [2,3]. In terms of clinical practice and research, the evaluation of severity and activity objectively in uveitis is an important and challenging issue. Therefore, the classi cation of uveitis based on the degree of in ammation is crucial for the support of follow-up examinations to evaluate the response to treatment in clinical practice and standardization of evaluation to maintain the quality of clinical research in different sites [4]. In 2005, the SUN working group developed grading schemes for the degree of in ammation in uveitis [5]. These grading schemes for the severity and activity of uveitis have been established in anterior chamber cells, anterior chamber are, and vitreous haze. However, the SUN working group revealed that no consensus could be reached on the grading system for vitreous cells [5].
Conversely, recent advances in optical coherence tomography (OCT) technology have enabled visualization of intraocular in ammatory cells as hyperre ective particles [6,7]. Moreover, in a mouse model of experimental autoimmune uveoretinitis (EAU), a spectral domain OCT (SD-OCT) scoring system was already used to evaluate the severity of ocular in ammation with intraocular cellular in ltration [8].
By contrast, to the best of our knowledge, there has been no grading to evaluate uveitis. Therefore, we attempted to evaluate vitreoretinal cellular in ltration with SD-OCT and proposed a new OCT-based grading system on vitreous cell in ltration in uveitis.

Patient characteristics
In 26 patients with posterior or panuveitis, the mean ages were 55.0 ± 19.1 yo with a male:female ratio of 9:17. In 40 eyes with uveitis, 26 (50%) eyes had granulomatous uveitis, and 14 (27%) eyes had nongranulomatous uveitis. Six (15%) eyes exhibited infectious uveitis, and 34 (85%) eyes demonstrated noninfectious uveitis. In 22 control subjects, the mean ages were 52.3 ± 17.1 with a male:female ratio of 10:12. Table 1 shows the causes of uveitis in the eyes included in this study. The major causes were sarcoidosis (n = 4) and Vogt-Koyanagi-Harada disease (n = 4). There were 16 eyes with unclassi ed intraocular in ammation.  Table 2 shows the sensitivity, speci city, and interobserver agreement of modalities in the diagnosis of active in ammation. Regarding sensitivity, OCT-based grading (grade ≥1) was as good as those of slit lamp examination and fundus photography (85% and 93%, respectively) and better than the presence of vitreous CLHPs on SD-OCT (67.5%). With respect to speci city, all modalities were good. The grade for vitreous cells by slit lamp examination and fundus photography was 100%, and that for the presence of vitreous CLHPs and OCT-based grading (grade ≥1) was 95.5%. In the interobserver agreement of modalities, the presence or absence of intravitreal CLHPs on SD-OCT and positive scores based on OCTbased grading (grade ≥1) were almost perfect (k = 0.97 and k=0.84, respectively), but those in fundus photography were substantial (k = 0.75). Table 3 shows the numbers of patients with the exact score classi ed by OCT-based grading in uveitis and vitreous haze classi cation. The interobserver agreement of the score in the OCT-based grade was moderate (κ = 0.5).

Reproducibility of OCT-based grading scores
Moreover, the score in the OCT-based grading in uveitis was signi cantly correlated with the score in the vitreous haze classi cation (R = 0.345, p = 0.029), as shown in Figure 2.

Discussion
In this study, we showed that SD-OCT could show vitreoretinal cellular in ltration in uveitis as CLHP and proposed a new grading system of the severity of in ammation in uveitis. The sensitivity and speci city were good in the presence of in ammation using the OCT-based grading (grade ≥1), and the interobserver agreements were comparable with standard vitreous haze classi cation. Moreover, this OCT-based grading of uveitis matched a clinical classi cation in the vitreous haze.
The SUN working group proposed the Grading Scheme for Anterior Chamber Cells and Flare as standard grading scales to evaluate in ammation in the anterior segment [5]. Conversely, a representative grading scale as a vitreous in ammation indicator has been vitreous haze at present, which was a complex indicator because the worsened vitreous haze was caused by the accumulation of some factors: in ammatory cells and proteins in the vitreous [9]. Therefore, when the score of vitreous haze changed, we must consider various factors that could affect it. The recent study of SD-OCT to measure vitreous haze was also reported [10]. By contrast, the SUN working group could not reach acceptance with the classi cation of in ammatory cells in vitreous.
Recently, as the OCT image resolution improves, some clinical studies reported that SD-OCT could capture even subtle intraocular in ammation [6,7]. Especially in toxoplasmosis, intraretinal hyperre ective particles around active lesions were detected even in time domain OCT. Moreover, OCT showed vitreous hyperre ective particles since SD-OCT was released [11,12]. Oré ce et al. showed that scattering hyperre ective particles in the vitreous were observed in 18 eyes of 24 eyes (75%) in patients with ocular toxoplasmosis [12].
In practice, Spectralis OCT, which we used as representative SD-OCT in this study, had the optical axial resolutions of 7 µm and digital axial resolutions of 3.5 µm/pixel. In a mouse model of EAU, SD-OCT matched clinical and pathological ndings in any severity of EAU. Previously, fundus photography imagebased grading score had been attempted to evaluate the severity of in ammation in the EAU mouse model. However, OCT-based grading scores via vitreoretinal imaging have been recently established based on its superiority of accuracy in measurement on SD-OCT, compared to fundus photography [8,13,14]. Especially, cellular in ltration has been shown to be modi ed according to the severity of ocular in ammation in EAU. Therefore, the number of cellular in ltrates and other OCT ndings could monitor the severity of ocular in ammation. It was shown that choroidal in ammation existed as a high re ectivity of the retinal pigment epithelium when SD-OCT images in EAU model mice met grade ≥1 of EAU.
In this study, classi cation was based on the spreading area and number of hyperre ective particles according to some previous studies, including the EAU mouse model. Grade 1 indicates a small number of CLHPs and CLHPs limitedly located on the retinal surface. We believe that CLHPs in this stage do not reach the vitreous cavity due to the low severity of in ammation. Grade 2 also indicates a small number of CLHPs, but CLHPs spread to the vitreous cavity because the severity of in ammation in grade 2 is higher than that in grade 1. Grade 3 indicates a number of CLHPs (11-50 particles) more than grade 2.
Grade 4 indicates a large number of CLHPs and diffuse spread in the vitreous cavity because of the higher severity of in ammation than any other grade.
Reproducibility is important to establish a standard grading system. The SUN working grading score was already evaluated with respect to the interobserver agreement. Kempen et al. reported that exact agreement by three uveitis specialists on the grading of AC cells was 0.34 to 0.43 (low to moderate agreement), that on the grading of AC are ranged from 0.50 to 0.64 (moderate agreement), and that grading of vitreous haze was 0.5 (moderate agreement). Moreover, the agreement range was 0.48 to 0.51 (moderate agreement) even when vitreous cells were graded as absent or present [15].
Regarding the interobserver agreement of modalities in this study, the presence or absence of intravitreal CLHPs on SD-OCT and that of positive scores based on OCT-based grading (grade ≥1) were almost perfect. They have higher reproducibility than fundus photography with a substantial agreement and the presence of vitreous cells by slit lamp examinations. Moreover, the exact agreement (κ = 0.5) of the OCTbased grading score was almost equal to those of any SUN grading score. Therefore, the OCT-based grading system could be acceptable in terms of reproducibility.
The present study had limitations in the study design due to its retrospective nature and a relatively small number of subjects. A prospective study with a larger sample size will be required to con rm the results of the present study. An additional limitation was that only Japanese patients were included and only Spectralis OCT was used in OCT imaging. However, our results suggested that SD-OCT could depict vitreoretinal cellular in ltration in uveitis as CLHP and provide a new grading system of the severity of in ammation in uveitis with good sensitivity and speci city to differentiate uveitis in the posterior segment and control subjects. This information might be essential to establish a new grading system using SD-OCT with a different evaluation concept from the SUN working scheme.
To conclude, this newly proposed OCT-based grading in vitreoretinal cellular in ltration can provide good sensitivity and speci city to differentiate uveitis involving the posterior segment and no ocular in ammation subjects and may be applicable in evaluating the severity of ocular in ammation.

Study design and participants
This is a retrospective comparative chart review focusing on vitreoretinal cellular in ltration using OCT.
All cases in this study were Japanese individuals recruited from the Department of Ophthalmology at the Kobe University Hospital in Japan. This study was approved by the Institutional Review Board at the Kobe University Graduate School of Medicine and conducted according to the Declaration of Helsinki. All subjects provided written informed consent.
The patients in this study initially visited the Kobe University Hospital between January 2016 and January 2017. Forty eyes of 26 patients with posterior uveitis or panuveitis and 22 eyes with no apparent abnormal ocular ndings from 22 individuals as sex-and age-matched controls during the same periods were included in the study. Patients with anterior or intermediate uveitis and cataract grade ≥3 in the Emery-Little classi cation and without available good-quality OCT images were excluded.
Fundus photographs that were obtained with 50° centered on the fovea were also collected. Fluorescein angiography and/or indocyanine angiography were performed as required.
All clinical data for this study were collected from clinical records, which included imaging at the phase with active in ammation before starting a new treatment or additional treatment in the Kobe University Hospital. According to vitreous haze grading, which was based on the scale developed by Nussenblatt et al. and subsequently de ned by the SUN working group [5], we also collected a vitreous haze score based on slit lamp examination or fundus photography from the clinical records.

Classi cation
In the evaluation of vitreal cellular in ltration, we selected vertical and horizontal cross-sectional SD OCT images through the fovea using B-scan averaging >50 images at the rst visit at the Kobe University Hospital. Next, a vertical or horizontal scan, on which a larger number of cells were found, was analyzed to evaluate vitreal cellular in ltration. Then, we evaluated the OCT-based grading (Figure 1). The SD OCT cross-sectional images through the fovea in all subjects were graded as follows: grade 0, absence of CLHP; grade 1, ≤10 CLHP on the retinal surface without vitreous CLHPs; grade 2, retinal CLHPs with ≤5-10 CLHPs in the vitreous; grade 3, 11-50 vitreous CLHPs; and grade 4, ≥50 vitreous CLHPs with diffuse distribution.
Subsequently, we selected the scan with a higher OCT-based grade score for statistical analyses. If horizontal and vertical scores were equal, we selected the vertical scan. W.M. and S.K. evaluated the OCT grading and the presence/absence of any abnormality on fundus photographs.

Outcomes
The primary outcome measures were the sensitivity and speci city of the OCT-based classi cation for detecting active intraocular in ammation. The sensitivity was calculated as the number of uveitis cases that were correctly classi ed with the presence of vitreous CLHPs on SD OCT image or that of positive scores based on OCT-based grading (grade ≥1), divided by all uveitis cases. Speci city was calculated as the number of control cases that were correctly classi ed, divided by all control cases.
The secondary outcome measures were the correlation of the OCT-based grading score with vitreous haze grading score and interobserver exact agreement of the OCT-based grading scores.

Statistics analyses
Interobserver agreement was evaluated using the κ method. Using Spearman's coe cient of rank correlation, bivariate correlations were analyzed. Statistical signi cance was de ned as a P-value of <0.05. All statistical analyses were conducted using MedCalc version 15.4 software (MedCalc Software, Mariakerke, Belgium), and a P-value of <0.05 was considered statistically signi cant. According to the descriptive system proposed by Landis    Correlation between OCT-based grade and vitreous haze grade in eyes with uveitis. Spearman's coe cient of rank correlation (R) is 0.345, and P-value is 0.029.