Abstract
Purpose :
Reduced thickness of the macular ganglion cell - inner plexiform layer (GCIPL) is indicative of early diabetic neurodegeneration. We therefore wanted to determine the comparability of GCIPL thickness measurements of 2 spectral-domain optical coherence tomography (SD-OCT) devices in patients with diabetic macular edema.
Methods :
Retrospective analysis of OCT scans in patients treated for diabetic macular edema. Macular cube scans with a signal strength ≥ 6 by Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA, USA) were compared to matched volume scans acquired by Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany) of the same examination date. Since Cirrus detects GCIPL thickness in an elliptical annulus centered on the fovea, ganglion cell and inner plexiform layer thicknesses of the inner ETDRS subfields were combined on Spectralis to obtain the area most comparable to this ellipsis. Automated segmentation was performed by Cirrus 7.0 and Heidelberg Eye Explorer 1.9.10.0 software versions. Due to Cirrus software limitations, manual correction of GCIPL segmentation was only possible for Spectralis scans and was applied when necessary.
Results :
Paired OCT scans of 20 eyes of 20 patients (mean age±SD 64±8 years, 6 female) were included in this comparative analysis. Average GCIPL thicknesses were 62±20 μm, 91±9 μm and 90±10 μm on Cirrus, automated and corrected Spectralis scans, respectively. GCIPL was significantly thicker on Spectralis (corrected or uncorrected) compared to Cirrus (p<0.001, respectively). Interdevice correlation (Cirrus versus automated Spectralis) resulted in a Spearman ρ=0.72. Bland-Altman plots showed a systematic difference between Spectralis and Cirrus measurements with a larger difference at thinner and smaller difference at thicker GCIPL values.
Conclusions :
GCIPL thickness obtained by different SD-OCT devices varied significantly and should therefore not be used interchangeably. Lacking interdevice agreement may be due to differences in the segmentation algorithms, the size of analysed areas and segmentation errors. As SD-OCT is an indispensable tool for identifying retinal GC loss and thinning of the inner retina as signs of early diabetic neurodegeneration, clinicians should be aware of this discrepancy when monitoring patients imaged by different OCT instruments.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.