Novel grading system for quantification of cystic macular lesions in Usher syndrome

Background To evaluate novel grading system used to quantify optical coherence tomography (OCT) scans for cystic macular lesions (CML) in Usher syndrome (USH) patients, focusing on CML associated alterations in MOY7A and USH2A mutations. Methods Two readers evaluated 76 patients’ (mean age 42 ± 14 years) data prospectively uploaded on Eurush database. OCT was used to obtain high quality cross-sectional images through the fovea. The CML was graded as none, mild, moderate or severe, depending on the following features set: subretinal fluid without clearly detectable CML boundaries; central macular thickness; largest diameter of CML; calculated mean of all detectable CML; total number of detectable CML; retinal layers affected by CML. Intra-and inter-grader reproducibility was evaluated. Results CML were observed in 37 % of USH eyes, while 45 % were observed in MYO7A and 29 % in USH2A cases. Of those with CML: 52 % had mild, 22 % had moderate and 26 % had severe changes, respectively. CML were found in following retinal layers: 50 % inner nuclear layer, 44 % outer nuclear layer, 6 % retinal ganglion cell layer. For the inter-grader repeatability analysis, agreements rates for CML were 97 % and kappa statistics was 0.91 (95 % CI 0.83-0.99). For the intra-grader analysis, agreement rates for CML were 98 %, while kappa statistics was 0.96 (95 % CI 0.92-0.99). Conclusions The novel grading system is a reproducible tool for grading OCT images in USH complicated by CML, and potentially could be used for objective tracking of macular pathology in clinical therapy trials.

Morphologically CML is characterized as a focal or generalized retinal thickening due to accumulation of fluid within retinal layers, which may lead to the formation of a lamellar or full-thickness macular hole. Moreover, even non-significant cystoid changes might contribute to a progressive impairment of visual acuity in RP, likely due to the degenerative process affecting retinal layers [20]. The central macula thickness (CMT) is assumed to be a potential morphological end-point for evaluating safety and efficacy of a treatment modality, which is aimed primarily for improving macular photoreceptor cells associated structure and/or function. Thus, it is of a key importance to have a sophisticated grading system for quantification of CML in USH-RP associated with CML.
Even though CML is a frequent complication of RP and despite of several attempts to grade CML [21,22], a need remains for an image grading protocol based on optical coherence tomography (OCT) to assist clinicians in describing all relevant aspects of an individual type of CML in USH-RP cases. We propose a new CML grading system for USH-RP patients, including comparative evaluation of CML between USH types and most frequent USH-associated genes, i.e. MYO7A and USH2A. Such a grading system would allow evaluation of USH-RP prognostic factors and treatment decisions in a very detailed way.

Subjects and database
This prospective observational cohort study was conducted according to the tenets of the Declaration of Helsinki and received authorization from Ile-de-France V ethical board committee for human research (June 4th 2013) and from the French regulation agency for medication (September 3rd 2013) (http://clinicaltrials.gov/ct2/ show/NCT01954953). The Eurush database (www.eurushdatabase.org) was used where subjects' data were uploaded on a prospective manner. The participants were recruited at the Centre Hospitalier National d'Ophtalmologie des Quinze-Vingts, Paris, France. The study procedure was fully explained to all patients, and their informed written consent was obtained. Seventy six patients with a clinical diagnosis of USH were enrolled (age 42 ± 14 years). The inclusion criteria were: clinical characteristics for USH1, USH2 and USH3 as defined by the Usher syndrome consortium [23]; informed consent and agreement to participate in the study. Exclusion criteria were: systemic pathologies or severe ocular pathologies; systemic or topical medication usage; otolaryngology pathologies which could contaminate the results; unwillingness to provide a blood sample for genetic test; unwilling and/or unable to undergo the study procedures. In more details we analyzed two groups, i.e. having mutations in genes USH2A and MYO7A.
Each patient underwent standard comprehensive ophthalmological examinations, which included assessment of the medical and family history, best corrected visual acuity (BCVA) using Early Treatment Diabetic Retinopathy Study (EDTRS), tonometry, slit-lamp biomicroscopy, direct and indirect ophthalmoscopy, static and kinetic perimetry using standard III4e isopters (Haag Streit, Koeniz, Switzerland). Full-field electroretinography (ERG) was performed with the Espion system (Espion Diagnosys, Littleton, MA), whereas multifocal electrophysiological recordings were carried out using the VERIS 4.9.1 system (Electro-Diagnostic Imaging, Inc., Redwood City). OCT and fundus autofluorescence (AF) were carried out using HRA II confocal scanning laser ophthalmoscope (Heidelberg Engineering, Heidelberg, Germany).
Standard cystic macular lesions grading procedure OCT image grading was performed in XV/XX reading centre (Paris, France). High quality macula scans were acquired using Spectralis HRA + OCT (Heidelberg Engineering, Dossenheim, Germany). The macular acquisition protocol consisted of 19 line raster fovea centered scans with at least a 15°by 15°region. The standard file transfer protocol was employed to transfer images in anonymous format to a local server. Five single scans lines, i.e. foveal scan, two below and two above the fovea were used for determining CML in both eyes. Both eyes were checked for CML. Only one (right) eye high quality cross-sectional image through the fovea was used for further grading. If CML were not present in foveal line scan but appeared in other 2 lines below or above the fovea, the closest line to the fovea was used for grading.
The trained graders performed OCT image grading using a novel grading system for CML quantification. In addition, 3 randomly selected tomograms were provided to each grader to test reproducibility. All reproducibility testing were performed 5 weeks after initial grading to avoid possible reader recall of the initial arbitration and to monitor temporal variability. The intra-grader and inter-grader reproducibility was calculated, in details described elsewhere [24]. Scan quality was estimated depending on presence/absence following features: fovea non determinable on OCT, incorrect positioning, and very poor saturation due to cataract and/or vitreous strands and opacities.
The protocol used to define the CML included the presence of cystoids spaces defined as hypo-reflective zone visible on at least two views of the sequential line scans in macula area. The boundaries of each CML were manually marked using Adobe Photoshop CS5.5 version. A grading protocol was defined covering all possible morphological alterations and characteristic patterns typically associated with CML, namely: 1. Subretinal fluid without clearly detectable CML boundaries (F s ); 2. CMT; 3. Largest diameter of CML (D n ); 4. Calculated mean of all detectable CML (D); 5. Total number of detectable CML (N); 6. Retinal layers affected by CML (L) (Fig. 1). Hence, CML was graded as mild, moderate or severe depending on above described features pattern (Table 1).
Currently, there is no available standard CML grading system and CMT is used for edema evaluation. Only for comparative reasons with the literature data, we extinguished CMT criterion from CML grading system for the more detailed analysis. Once the grading database was verified and closed, the grading and clinical data were merged for the analysis.

Statistical analysis
Statistical analysis was carried out using SPSS software (version 19.0, SPSS Inc., Chicago, USA). We analyzed CML in the whole cohort and in two groups associated with mutations in USH2A and MYO7A. Pearson correlation (r) was used to assess the correlation between central macular thickness and age in a whole cohort and associated genes, namely USH2A and MYO7A. One-way analysis of variance (ANOVA) was performed to detect differences between CML severity in associated gene USH2A and MYO7A. The percent agreement and the kappa statistics were calculated as measures for categorical variables (quality of scan and presence of CML). Percent agreement was computed as the number of concordant grading pairs divided by the total number of grading pairs multiplied by 100. All numerical data are presented here as arithmetic means ± standard deviation (SD). A two side P value less than 0.05 was considered statistically significant.

Main clinical characteristics
Our cohort consisted of 76 patients, 41 (54 %) were males. All 76 (100 %) patients were found to carry at least one mutation, while 68/76 (89 %) of them carrying two mutations in USH-associated genes. The methodology used to detect mutations (complete exon sequencing) in USH associated genes is described in details elsewhere [2]. The prevalence of mutations in USH-associated genes was as following: 13 MYO7A patients tended to have worse BCVA compared to USH2A, although the difference did not reach statistically significant level (p > 0.05).

Quality assurance data for a novel grading system
The evaluation of the proposed system is done using Eurush database estimating intra-grader and inter-grader repeatability. The time to grade one tomogram with a b  Fig. 2 The prevalence of cystic macular lesions severity in Usher syndrome cases. The prevalence of cystic macular lesions in USH 1 and USH 2 a, the prevalence of cystic macular lesions in USH2A and MYO7A b CML required in averaged 5 min ± 1 min for the first and 6 min ± 1 min for the second grader (not statistically significant p > 0.05). The difference in mean CML diameter was 2 μm ± 1.5 μm between the two graders performed grading of OCT scans. Reproducibility rates were high for the categorical variables scan quality at the fovea and CML. For the inter-grader repeatability analysis, agreements rates were 97 % (95 % CI 0.915-1.00) and 95 % (95 % CI 0.89-1.00) for the scan quality at the fovea and CML, respectively. The kappa statistics was 0.93 (95 % CI 0.85-1.00) and 0.91 (95 % CI 0.83-0.99) for the scan quality at the fovea and CML, respectively. For the intra-grader analysis, agreement rates were 96 % (95 % CI 0.80-1.00) and 98 % (95 % CI 0.89-1.00) for the scan quality at the fovea and CML, respectively. Likewise, kappa statistics was 0.91 (95 % CI 0.80-0.95) and 0.96 (95 % CI 0.92-0.99) for the scan quality at the fovea and CML, respectively.

Discussion
In the present study, we established a novel grading system that showed to be reliable and reproducible in our reading centre for evaluating CML in USH-RP patients. Our system consists of CML detection assessed by trained graders, followed by evaluation of CML severity and assessment of affected retinal layers. Using the novel grading system OCT scans can be graded in a standard manner in about 5 minutes.
To the best of our knowledge, there are limited references in the literature regarding how to evaluate CML severity in USH-RP cases. We found that using our standard method, CML quantification gives a detailed view on retinal morphological changes which is crucial for disease progression monitoring and exact end-points estimation. In addition, the incidence of CML was compared within the most frequent USH-associated genes, i.e. MYO7A and USH2A. In our cohort of 76 USH patients, CML rate was found to be 37 %, while in MYO7A was of 45 % and USH2A was of 29 %. Almost half of MYO7A cases were associated with CML, although a statistically significant difference between MYO7A and USH2A groups was not reached. In addition, MYO7A cases tend to represent by severe CML, while USH2A tend to have mild CML. So far no pathophysiological studies were performed analyzing CML incidence differences in USH2A vs MYO7A. Thus, it definitely requires more detailed investigations.
Our findings of 37 % having CML in OCT scans are comparable to Hajali et al's, namely 35 % USH2 and 39 % autosomal recessive RP patients had CML in OCT scans [25,26]. It must be acknowledged that there is a great difference in the incidence of CML among the various cohorts of RP patients, ranging from 8-15 % [12,16,17] to 25-56 % [14, 15, 19, 25]. One reason might be that ascertaining CML might vary due to race and ethnicity differences. CML incidence rates are highly dependent on RP inheritance pattern (recessive, dominant or X linked) and so genotypic preponderance can also influence the results [11]. Some studies were very small which can also lead to unstable estimates. Finally, the lack of structured grading might also make ascertaining CML difficult and unreliable.
In the current study we showed that eyes with CML have 30 % greater CMT (235 ± 98 μm) compared to those without CML (164 ± 59 μm). Gorovoy et al. [20] also reported similar findings: 28 % of RP patients had CML and half of them had increased central foveal thickness (239 ± 10 μm). Some differences could be explained by differences in age between studies and the variability in disease stage, in particularly the presence of retinal atrophic changes. Our study clearly showed that younger patients tended to have higher rates of CML (r = −0.47, p = 0.01). In a large enough cohort, multiple regression analysis where CML severity could be tested against age and CMT would potentially provide new insight into the natural disease history.
Additionally, we showed that CML tended to be present in INL (50 %) and ONL (44 %). Makiyama et al. [18] showed 27 % incidence of CML, while cystic lesions were observed in the same layers as in our study, namely INL (99 %) and ONL (28 %) and a minority in RGC. Fakin et al. [19] observed 56 % CML in USH-RP patients, while CML were distributed in INL (60 %), INL and ONL (33 %) and OPL (7 %). To the best of our knowledge, there was no pathophysiological studies performed so far analyzing CML affinity for the particular retinal layers.
The natural course of CML associated with RP is incompletely understood with several theories proposed. Heckenlively et al. [26] proposed that CML might be a result of an autoimmune process. The high incidence of CML in INL and ONL could be explained by microglial cells activation as a result of the autoimmune process  [27 -29]. An alternative hypothesis could be that CML is a result of a breakdown in the retinal barrier at the level of the retinal pigment epithelium which leads to a fluid leakage in the macula. The breakdown of bloodretinal barrier potentially results in accumulation of plasma proteins which exert a high oncotic pressure in the neural interstitium, which tends to produce interstitial edema [30][31][32]. It is possible that both mechanisms occur together [33,34]. Lastly, it is possible that the "cystic spaces" are actually due to apoptosis and therefore simply represent a space left from tissue loss [35].

Conclusion
A standard OCT-based retinal image evaluation in USH-RP represents an urgent need both for clinicians and scientists and calls for a reading centers focused of OCT evaluation. The proposed grading system supplies a reproducible tool for CML quantification in USH-RP, and eventually could be used for objective tracking of macular pathology in clinical trials. Using our grading system we observed that CML is a common retinal complication in USH-RP and showed that CML tended to be present in both inner and outer nuclear layer. Moreover, we found that MYO7A cases tend to have a higher CML incidence rate and more severe CML compared to USH2A.
We describe for the first time a novel grading system for CML evaluation that is based on expert knowledge and requires trained personnel. In the future, the novel grading system might be applied using semi-automatic image analysis for CML severity grading, taking into account functional and/or other significant structural measures [36][37][38][39][40]. The system has a potential to be implemented in reading centers as a sophisticated tool for retinal structure evaluation in USH-RP. Surely, it would be directly applicable to any CML in other retinal diseases (e.g. diabetic retinopathy, age related macula edema) and aid treatment challenges in other rare retinal degenerations.