Multidimensional Functional Phenotyping Based on Photoreceptor-Directed Temporal Contrast Sensitivity Defects in Inherited Retinal Diseases.

Purpose: To identify patterns of functional defects in perifoveal photoreceptor-directed temporal contrast sensitivities (tCSs) in patients with inherited retinal diseases.

Methods: We retrospectively studied patients with RP1L1-associated occult macular dystrophy (OMD), Stargardt disease (STGD), and RP. Photoreceptor-directed tCS directed at L-, M-, S-cones and rods at different temporal frequencies were measured using a four-primary LED-stimulator with an annular test field (2° inner diameter and 12° outer diameter). Mean defects (MDs) were calculated by subtracting sensitivities from age-correlated normal values and averaging defects in frequency ranges where single postreceptoral pathways mediate flicker detection. Each patient was characterized by 6 MD values (one value each for S-cones [SMD] rods [RMD]; two values each for L- [LMDlow/high] and M-cones [MMDlow/high], where low refers to 1-6 Hz and high to 8-20 Hz temporal frequency ranges). Groups of similar phenotypes were identified with (supervised) decision trees and (unsupervised) hierarchical classification trees (based on nearest neighbors) and compared with the clinical diagnoses.

Results: The pruned decision tree used RMD for separating RP/STGD from normal/OMD, LMDlow for separating OMD from normal, and SMD for discriminating between RP and STGD. The accuracy was 66%. The hierarchical tree (independent of clinical diagnosis) was cut to four clusters, resulting in one cluster containing mainly normal participants, one cluster with severe L- and M-cone defects caused by OMD or STGD, one cluster with severe rod defects (4/5 with RP) and a large cluster with intermediate rod and cone defects that was dominated by RP and STGD patients.

Conclusions: LMDlow, SMD, and RMD were the most important parameters. Photoreceptor-directed tCSs allow sophisticated functional phenotyping of inherited retinal diseases and complement other structural and functional parameters for genotype-phenotype correlations.