Clinical and artificial intelligence characterisation of X-linked retinoschisis for gene therapy

Thesis

X-linked retinoschisis (XLRS) is a monogenic inherited retinal disease (IRD) caused by pathogenic variants in the RS1 gene and represents a frequent cause of early-onset visual impairment in males. The disease is characterized by intraretinal cystoid cavities (ICC), retinal layer splitting, and progressive structural and functional changes. Despite increasing interest in gene therapy for XLRS, the natural history of the condition, optimal therapeutic window, and appropriate anatomical and functional outcome measures for interventional trials remain incompletely defined.

The aim of this thesis was to characterise the natural history of XLRS, evaluate structural and functional parameters, and explore the application of artificial intelligence (AI)based image analysis to support endpoint selection for future gene therapy studies. Multicentre retrospective analyses demonstrated that the XLRS phenotype is dynamic and evolves over time, with findings consistent with a therapeutic window extending into the third decade of life. Retinal nerve fibre layer thinning was observed in three quarters of patients and was topographically associated with macular and peripheral retinoschisis. Marked intrafamilial phenotypic variability was identified among patients carrying identical RS1 variants, and the phenotypic spectrum of XLRS was expanded to include cases with minimal macular involvement. 

An AI-based algorithm was developed to quantify ICC volume on optical coherence tomography across heterogeneous acquisition protocols. ICC volume emerged as the most suitable anatomical outcome measure for future XLRS gene therapy trials, while retinal sensitivity assessed by microperimetry represented the most sensitive functional endpoint, outperforming conventional retinal thickness parameters and best-corrected visual acuity. These findings support the use of disease-specific, morphological-functional aligned endpoints in XLRS.

From a translational perspective, the results inform key aspects of clinical trial design, including endpoint selection, age at intervention, and the use of the fellow eye as a control. They further highlight disease-specific considerations relevant to surgical delivery and safety assessment in XLRS gene therapy. Overall, this work provides a comprehensive basis for natural history assessment and outcome measure selection with the aim of improving the design and interpretability of future XLRS gene therapy trials.

Authors: Kiraly, P

• DOI: 10.5287/ora-agbkqng8p
• Type of award: D.M
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English