Inter-device reliability of swept source and spectral domain optical coherence tomography and retinal layer differences in schizophrenia

Abstract

Introduction

Optical coherence tomography (OCT) is used to study retinal structure in schizophrenia. Changes in retinal structure, especially the retinal nerve fiber layer (RNFL) has been correlated with psychotic disorders. Measurement variability is a concern since there are various generations of OCT devices. We investigated the inter- and intra-device agreement of macular thickness between spectral domain (SD−OCT) and swept source−OCT (SS−OCT), and compared macula and peripapillary group differences in schizophrenia using SS−OCT.

Methods

Macular OCT thickness was obtained for schizophrenia (SZ, n = 30) and healthy controls (HC, n = 22) subjects using SD−OCT (Heidelberg Spectralis) and SS−OCT (DRI Topcon Triton). Peripapillary thickness was obtained using SS−OCT. RNFL, ganglion cell-inner plexiform complex (GCL+), RNFL plus GCL+ (GCL++), and macular thickness were collected. Clinical and cognitive data were gathered. All statistical analyses were performed using R software.

Results

There was excellent inter-scanner agreement for GCL + and GCL++ with ICC’s between r = 0.92−0.99. Good-to-excellent intra-scanner (OD vs OS) agreement was present except for macular RNFL in the SS−OCT device. No significant peripapillary group differences were identified. Poorer GAF scores were correlated with thinner macular layers and thinner overall peripapillary retinal layer. Greater mania symptoms were associated with smaller peripapillary GCL + thickness (r=-0.43, p = 0.03). Poor overall cognition (Brief Assessment of Cognition in Schizophrenia) was associated with smaller overall peripapillary retinal thickness (r = 0.36, p = 0.02).

Conclusion

While there is RNFL variability, GCL + and GCL++ are comparable between scanners SD−OCT and SS−OCT. Given that RNFL thinning is strongly implicated in psychotic disorders, the use of OCT scanners should not be interchanged due to increased RNFL measurement variability.