The current method for assessment of the inflammatory status in eyes with uveitis is almost entirely based on subjective clinical estimates according to standardized scales for both anterior and posterior segment. In the case of vitreous inflammation, the gold standard has been the National Eye Institute (NEI) system for grading of vitreous haze, often known as the Nussenblatt scale. This classification is based on estimation of the clarity of the fundus when viewed with the indirect ophthalmoscope and a 20 diopter lens in comparison to a set of reference photographs. The NEI vitreous haze score has been approved as a surrogate endpoint by the United States Food and Drug Administration and is used as a primary outcome measure in the majority of clinical trials in uveitis. Vitreous haze score does, however, have a number of limitations: it is subjective, with only moderate interobserver agreement ; it is noncontinuous, grading disease activity in large steps between vitreous haze score categories; and it is poorly discriminatory at the lower levels of vitreous inflammation that represent the majority of patients with active vitreous inflammation (ie, +0.5, +1 vitreous haze score) and therefore arguably a challenge for clinical assessment, particularly in clinical trial scenarios.

There is consensus in the uveitis community on the need for objective measures of inflammatory activity in sight-threatening uveitis. With this aim, a recent proof-of-concept study demonstrated that spectral-domain optical coherence tomography (SD OCT) images could be processed to obtain objective measurements of vitreous inflammation in eyes with intermediate uveitis, posterior uveitis, and panuveitis. This method was based on the determination of the signal intensity of the vitreous compartment, which was then compared to that of the retinal pigment epithelium (RPE) to generate an optical density ratio with arbitrary units that showed good correlation with vitreous haze score, but further validation was required. First, the original study was too small to allow subgroup analysis of common patient factors that might render the technique unreliable, notably phakic status or previous vitreoretinal surgery. Second, the original study employed an OCT device from a single vendor (Spectralis OCT; Heidelberg Engineering, Heidelberg, Germany), raising the possibility that the method might not be applicable across other medical systems. Third, these preliminary findings needed to be confirmed in a different patient cohort, using a different team of OCT technicians and independent analyzers to ascertain whether the method has applicability to and may be adopted by multiple end users.

To these ends this study aims to evaluate the previously proposed OCT-derived vitritis quantification method in a larger series of intermediate uveitis, posterior uveitis, and panuveitis eyes, using an alternative spectral-domain OCT platform in a geographically and demographically distinct study population. Such data would deliver an independent evaluation of an ability to obtain objective, continuous, and reproducible measurements of vitreous inflammation. A further aim is to determine the influence of phakic status or previous vitreoretinal surgery on these objective measurements.

Methods

All OCT image sets were obtained from patients attending a tertiary referral uveitis clinic (A. Adan) at Institut Clínic d’Oftalmología (ICOF), Hospital Clinic, Barcelona, Spain. Patients included in the study had intermediate uveitis, posterior uveitis, or panuveitis of different etiologies, with varying degrees of vitreous inflammation, and corresponding OCT image sets captured during routine clinical care, from a 6-year period (November 11, 2009-June 2, 2014). This study was approved by the Ethics Committee of the Hospital Clinic, Barcelona and was conducted in accordance with the Declaration of Helsinki.

Clinical Data

Demographic data collected from patients in the study included age, sex, uveitis anatomic location, uveitis etiology, current treatment, and any history of previous intraocular surgery. Clinical characteristics of study eyes collected include the following: (1) best measured visual acuity; (2) presence of keratic precipitates; (3) presence of posterior synechiae; (4) phakic status, classified as (a) phakic, (b) pseudophakic, and (c) aphakic; (5) anterior chamber (AC) activity; and (6) vitreous haze score using standardized protocols according to the NEI and Standardization of Uveitis Nomenclature (SUN) guidelines. All clinical data were collected during routine clinical care in an electronic medical records system and extracted for analysis. Only eyes with complete information of the above fields and corresponding OCT images were included in the analysis ( Supplemental Figure 1 ; Supplemental Material available at AJO.com ).

Optical Coherence Tomography Image Acquisition Protocol

All SD OCT image sets included in this study were acquired using a spectral-domain OCT system (Cirrus HD-OCT; Carl Zeiss Meditec, Dublin, California, USA) with the standard “Macular Cube” protocol. The Macular Cube protocol consists of 128 horizontally oriented B-scans acquired in a continuous, automated sequence and covers a 6 mm × 6 mm area. Each B-scan is 6 mm in length and composed of 512 equally spaced transverse sampled locations. The enhanced depth imaging mode was not used in any case, and the point of maximum sensitivity or zero delay line was maintained in the vitreous side. For the purposes of this study, only scans centered in the fovea were analyzed.

Qualitative Analysis of Optical Coherence Tomography Images

SD OCT images were qualitatively analyzed to assess the presence of (1) hyperreflective dots, larger and with greater density than background speckle noise as surrogate marker of cellular infiltrates into the vitreous ; (2) presence of epiretinal membrane preventing adequate transmission of light to the retinal pigment epithelium; and (3) severe anatomic disruption of retinal integrity and outer retinal layers/RPE status, preventing adequate delineation of the retinal pigment epithelium compartment of interest for the quantitative analysis. All cases with (2) and (3) were excluded from the subsequent quantitative analysis. Examples of clinical cases excluded are described in Supplemental Figure 2 (Supplemental Material available at AJO.com ).

Quantitative Analysis of Optical Coherence Tomography Images

Raw SD OCT images were exported from the Cirrus HD-OCT system and imported into OCTOR (Doheny Eye Institute, Los Angeles, California, USA), custom grading software that allows manual delineation of boundaries that define the compartments of interest. In each case, boundaries were manually segmented as per a previously described protocol for 3 B-scans passing through the foveal central subfield (central subfield of the Early Treatment Diabetic Retinopathy Study chart) by an experienced grader (J.Z.V.) masked to clinical vitreous haze score values. The boundaries were as follows: (a) vitreous top, the uppermost extent of vitreous space included in the OCT image; (b) internal limiting membrane (ILM), the inner boundary of the neurosensory retina; (c) RPE-inner limit, the inner boundary of the RPE; and (d) RPE-outer limit, the outer boundary of the RPE. These boundaries delineate the following spaces: (1) vitreous, defined as the space between the vitreous top and ILM (except cases with incomplete posterior vitreous detachment, where the innermost limit of the vitreous hyaloid was chosen, as including this retrohyaloid space filled with optically hyporeflective aqueous may underestimate the absolute intensity of the compartment); and (2) RPE, defined as the space between the RPE-inner and the RPE-outer boundaries ( Figure 1 ). The intensity of all pixels included in each of these compartments was summed to generate a corresponding mean OCT intensity value for the total scanned area, described as “VIT-absolute intensity” and “RPE-absolute intensity,” respectively. Finally, an arbitrary ratio was generated with these 2 values and expressed as “VIT/RPE-relative intensity” (VIT-absolute intensity/RPE-absolute intensity).

Segmentation protocol for quantitative analysis of the vitreous with OCTOR software. Spectral-domain optical coherence tomography (SD OCT) B-scan obtained from a 54-year-old patient with Vogt-Koyanagi-Harada disease–related panuveitis and moderate vitreous haze (clinical vitreous haze score +2). (Top) Raw SD OCT B-scan. (Middle) Boundaries segmented manually with OCTOR (top white line, arrowhead = vitreous top; grey line, asterisk = internal limiting membrane; large white arrow, white line = retinal pigment epithelium [RPE]-inner boundary; small white arrow, white line = RPE-outer boundary). (Bottom) Segmentation of the vitreous and the RPE compartments defined by the above boundaries (vitreous = gray, retina = light grey, RPE = dark grey). Vitreous signal intensity/RPE (VIT/RPE)-relative intensity of 0.4146.

Statistical Methods

Descriptive and frequency statistics were used to assess qualitative variables. Normality of quantitative variables (ie, VIT/RPE intensity ratios) was examined using histograms. To assess differences in VIT/RPE mean intensity between study groups, nonparametric Mann-Whitney U test and Kruskal-Wallis test were employed for 2-to-2 and ≥3 groups comparisons, respectively. Correlation between VIT/RPE intensity ratios and vitreous haze score was analyzed using nonparametric Spearman test, and a multivariable regression model was applied to assess the effects of clinical characteristics on SD OCT intensity values. Visual acuity (VA) measured in Snellen notation was converted to logMAR (logarithm of the minimal angle of resolution) equivalents for the purpose of statistical analysis. VA values recorded as counting fingers (CF), hand movements (HM), and perception of light (PL) were converted to 2.1, 2.4, and 2.7 logMAR, respectively. A P value of less than .05 was considered statistically significant. All statistical analysis was performed using SPSS 15.0 software (SPSS Inc, Chicago, Illinois, USA).