Introduction

Retinal thickness parameters as assessed by optical coherence tomography (OCT) have been used extensively for clinical trial eligibility, clinical trial endpoints, and clinical management of retinal diseases.1, 2, 3, 4, 5 However, the correlation between OCT-measured central retinal thickness and visual acuity (VA) is only modest among a variety of retinal diseases.6, 7, 8, 9, 10 Emerging literature suggests that both qualitative and quantitative features of the outer retina may correlate with VA in various macular pathologies, including retinal vein occlusion (RVO).9, 11, 12, 13, 14 If so, outer retinal morphology may be an important predictor of visual function and emerge as an important endpoint in clinical trials of RVO and other conditions.

Spectral domain OCT (SD OCT) technology has led to progress in outer retinal layer segmentation to enable specific layer measurements, such as photoreceptor length and outer segment length, although the algorithms have been thus far applied to eyes with no known disease or a few selected pathologies.15, 16, 17, 18, 19, 20 Controversy still exists in the identification of precise outer retinal layers.18, 21 The disruption of retinal architecture in pathology poses a challenge to the segmentation algorithm, which could result in erroneous thickness measurements.22 Integrity of the photoreceptor inner segment–outer segment (IS–OS) junction has been shown to correlate well with VA in cross sectional studies of patients with diabetic macular oedema, RVOs, and neovascular age-related macular degeneration.12, 13, 14, 23, 24 Longitudinal evaluation in large data sets would help confirm the validity of this variable.

The Standard Care vs COrticosteroid for REtinal Vein Occlusion (SCORE) Study employed high resolution time domain OCT (TD OCT; Stratus, Carl Zeiss Meditec, Dublin, CA, USA) images of eyes with macular oedema associated with RVO. The SCORE Study results showed only a moderate correlation between change in VA and change in OCT-measured center point thickness (CPT).25, 26 The purpose of this SCORE ancillary study is to evaluate the correlation between the IS–OS status and concurrent VA, and also the relationship between change in IS–OS status and change in VA in a longitudinal set of high resolution TD OCT images of patients with RVO.

Materials and methods

The SCORE Study design and baseline characteristics of the SCORE Study participants have been published previously.1 OCT scans were obtained using the Stratus OCT3 machine by certified imaging technicians at baseline and every 4 months thereafter for 36 months. The scanning protocol included the fast macular thickness scan protocol (128 A scans/B scan) and the high resolution cross hair scan protocol (512 A scans/B scan). The SCORE Study reading centre utilized paper prints of the scans for evaluation. However, the evaluation protocol for outer retina required gray scale images for identifying the IS–OS junction and photoreceptor layer. To accommodate this, the OCT scans included in the current study were evaluated in the manufacturer’s review software (Carl Zeiss Meditec). As the project involved retrieving digital OCT scans, evaluation was limited to scans from a single SCORE Study site only (University of Wisconsin). The high resolution cross hair scans were utilized as the resolution of fast macular scans was insufficient for evaluating the IS–OS junction.

Cross hair scans were viewed in the manufacturer’s review software using the align scan function with images displayed in gray scale. Scan centration was assessed and the length of the B scan corresponding to the central subfield was identified. Using both the horizontal and vertical cross hairs, the IS–OS junction in the central subfield was evaluated using a three step scale: normal, abnormal, and absent (Figure 1). ‘Normal’ IS–OS junction indicated a well-delineated hyper reflective signal passing through the entire length of the central subfield. ‘Abnormal’ IS–OS junction was defined as disruption in the signal intensity so that it appeared speckled or irregular, yet still visible. An ‘absent’ IS–OS junction indicated that the layer was not visualized. The lack of visualization could be due to the absence owing to pathology or due to blocked signal (haemorrhage, large cyst) or change in orientation of photoreceptors (subretinal fluid) or poor signal quality. Change in IS–OS status from baseline to month 12 was categorized as improved, no change, or worsened. Improved IS–OS status was defined as a change in IS–OS status from absent to abnormal, absent to normal, or abnormal to normal. Worsening of IS–OS status indicated that eyes changed from abnormal IS–OS at baseline to absent IS–OS at month 12.

Figure 1
figure 1

Evaluation of the photoreceptor inner segment–outer segment (IS–OS) status using the three step scale; left normal IS–OS, centre abnormal IS–OS and right absent IS–OS.

Full size image

The assessment was made separately for the two cross hair scans and the worse grade was selected for the eye. The reproducibility of the three step scale was assessed on a pilot data set evaluated by two independent graders (separate from the SCORE Study data set; n=30 eyes). Agreement for evaluation of IS–OS status was 80.6% (κ 0.71). All images thereafter were graded independently by a single evaluator. Each visit scan was graded independently without comparison with previous or baseline scans.

Photoreceptor length was measured as the distance between the outer plexiform layer (OPL) and the retinal pigment epithelium. This measurement was taken at three points in the central subfield of each cross hair scan using the software calipers, at the center point and 500 μm on either side of the center point. Both the OPL and the retinal pigment epithelium had to be distinctly visible to position software calipers on at least one of the three points on a B scan without which the measurement was considered ungradable. All available measurements were used to obtain a mean photoreceptor length in the central subfield.

Measurement of the best corrected VA was performed using the electronic Early Treatment Diabetic Retinopathy Study VA method at 3 m by a SCORE Study-certified VA examiner.

Statistical analysis was performed using the Kruskal–Wallis test to compare the distribution of VA and CPT across the three step scale (Wilcoxon rank-sum test when only two steps were available). As the data set was limited to a single site, analysis across treatment arms was not performed.

Results

OCT scans of 42 subjects enrolled at a single site were evaluated. At baseline, the IS–OS junction was absent in 30 (71%) eyes, abnormal in 12 (28%) eyes and normal in none. The 30 eyes with absent IS–OS junction were almost equally divided between central and branch RVO (Table 1). Retinal thickness at the centre point was greater and VA tended to be worse in these eyes than in those in which the IS–OS junction was classified as absent at baseline (Table 2). At month 12, eyes with absent IS–OS status had the worst VA, eyes with abnormal IS–OS status had intermediate VA, and eyes with normal IS–OS status had the best VA (P<0.01). The trend in the distribution of VA letter score between the three categories was also seen at the month 4 and 8 follow-up visits.

Table 1 Baseline characteristics of eyes with absent and abnormal IS–OS status
Full size table
Table 2 Comparison of IS–OS status at baseline and month 12 with other baseline parameters (n=42)
Full size table

Table 3 displays change in VA letter score for eyes with improved or no change in IS–OS status. There was a significant difference in the mean change in VA letter score between eyes with improved IS–OS and eyes with no change in IS–OS (+6.0 and −7.8 letters, respectively; P=0.03). There was no significant difference in the proportion of eyes with a gain in VA letter score of ≥15 letters between the two categories, although there was a trend (P=0.06). There was no difference in the distribution of mean change in CPT between the two categories (P=0.8). A similar analysis at month 24 showed no significant difference in change in VA letter score with change in IS–OS status; eyes that demonstrated an improved IS–OS status had a mean VA gain of 1.69±26.9 letters compared with a mean VA loss of 2.3±15.7 letters in eyes that showed no change in IS–OS status (P=0.29) (data not shown). In addition two eyes developed significant cataract at month 12 visits. The results did not change after excluding the two eyes.

Table 3 Change in IS–OS status between baseline and month 12 and change in visual acuity and center point thickness
Full size table

Table 4 cross-classifies IS–OS status at baseline and at month 12 and presents mean baseline to month 12 change in CPT and in VA. Decrease in CPT was greater in eyes with absent IS–OS junction at baseline than in those with abnormal IS–OS junction. Change in VA appeared to show a trend with change in IS–OS status. In 16 eyes that remained with absent IS–OS status at 12 months, the mean VA decreased by 10.4 letters, in 6 eyes with improvement in IS–OS status from absent to abnormal mean VA remained essentially unchanged (−0.8 letters), and in 8 eyes with improvement from absent to normal IS–OS junction mean VA improved by 8.9 letters. Results were similar for the 12 eyes with abnormal IS–OS junction at baseline, but not nominally significant, perhaps because of small numbers.

Table 4 Change in retinal thickness at center point and change in visual acuity between baseline and month 12 by IS–OS status
Full size table

Photoreceptor length was ungradable in 27 (64%) eyes. Visibility of OPL was a requirement for this measurement and this layer was disrupted in most eyes because of intraretinal fluid. Although the numbers were small, there was no significant correlation between photoreceptor length and IS–OS status at both baseline and month 12.

Discussion

In this cohort of eyes with macular oedema because of RVO from the SCORE Study, IS–OS status evaluated with TD OCT scans showed a significant association between change in IS–OS status and change in VA from baseline to 1 year. Independent of baseline category, IS–OS status returning to normal was associated with improvement in VA, and IS–OS in the absent category was associated with the loss of VA. Although there seemed to be a significant relationship between change in IS–OS status and change in VA, the range of VA letter score was wide. Interestingly, eyes that remained in the absent IS–OS category at month 12 did not remain stable in their VA, but instead showed a mean loss in VA letter score.

At baseline, IS–OS status was significantly associated with OCT-measured CPT, but not with VA and photoreceptor length. There were no cases with normal IS–OS status at baseline, indicating that outer retinal status is often affected in cases where there is VA impairment associated with macular oedema due to RVO. At month 12, IS–OS status was significantly associated with VA and OCT-measured CPT, but not with photoreceptor length. The association between thick retina and absent IS–OS junction at both baseline and follow-up visits raises the possibility that thickened retina confounds IS–OS assessment. Despite this potential confounder, change in IS–OS status between baseline and 12 months had a stronger correlation with change in VA than change in central retinal thickness (Table 3).

Image quality issues were common and may have confounded assessment, particularly at baseline. The IS–OS junction is not a true tissue interface and possibly represents a change in refractive index between the inner and outer segments of the photoreceptors.27 It is difficult to distinguish loss of IS–OS junction due to photoreceptor damage from absence of IS–OS junction due to various imaging artifacts. Overlying cysts haemorrhage and hard exudates may attenuate the reflectivity of deeper retinal layers, thereby decreasing the sensitivity of the classification of outer retina morphology. At baseline, the IS–OS status may have been misclassified as absent, not because of a structural abnormality but because of poor imaging of the layer. By month 12, the cysts, subretinal fluid and intraretinal haemorrhage had resolved in most eyes and, thus, were not an impediment to imaging the outer retina. At this time, we are unable to distinguish between absence of the IS–OS due to obscuration by artifact and absence of the layer due to pathology.

Recent studies have found correlations between IS–OS status and VA in various diseases in TD OCT and more recently SD OCT.12, 13, 14, 23, 24, 28, 29, 30, 31, 32, 33 Recent publications on using SD OCT further support the view that outer retinal integrity is an important factor for visual outcome in eyes with RVO.32, 33 However, there is lack of long-term follow-up in a prospective standardized fashion using SD OCT. The current ancillary study was performed to assess the relationship between change in IS-OS and change in VA status in a randomized clinical trial with a standardized visit schedule, imaging protocol, and VA measurements. Although high resolution TD OCT scans (512 A scans/B scans) with resolution similar to some SD OCT macular volume scan (‘cube’) protocols were used for evaluation, only two B scans were available in the cross hair protocol of the TD OCT in contrast with SD OCT where many more B scans are available for assessment of IS–OS status. Agreement between Topcon 3D OCT scans and Stratus cross hairs for status of the IS–OS junction in the central subfield using the three step scale was 80.6%, κ 0.7 (unpublished data). Thus, the results from this study using TD OCT are likely applicable to SD OCT.

Recent literature has focused on outer segment length as an important predictor of VA.8, 9 We did not analyze outer segment length in this sample due to the lack of tools for the required minute measurements in the manufacturer’s software. Photoreceptor length has also been found to be correlated with VA in macular oedema from RVO (r=−0.571)14 but in our experience is difficult to define in eyes with macular oedema due to RVO.

This SCORE ancillary study suggests that IS–OS status is positively correlated with VA and is a promising assessment that warrants further investigation. However, this variable is susceptible to artifacts and must be interpreted with caution. A comprehensive outer retinal morphology assessment including status of IS–OS, photoreceptor length and outer segment length measurements may be a useful endpoint in clinical trials.