Impact of decentration of SDOCT volume scans on retinal thickness measurements in diabetic macula edema in a phase III multicenter randomized controlled trial (RELATION)

Abstract

Purpose: To assess the impact of decentration of SDOCT volume scans on automated retinal thickness measurements in eyes with diabetic macula edema (DME) and the value of manual centration of the ETDRS grid by a reading center from a clinical trial perspective.

Methods: SDOCT volume scans from the RELATION study were analyzed by a reading center. In the RELATION study, patients with visual impairment due to DME were randomized to a combined group (ranibizumab plus central laser) and a laser monotherapy group. The study was terminated early due to approval of ranibizumab for DME. Following a manual correction of segmentation errors, foveal central subfield (FCS) and foveal center point (FCP) thickness values were calculated before and after manual centration of the ETDRS grid to the fovea. Retinal thickness, ETDRS visual acuity (VA) and SDOCT scan position were evaluated for their association with measurement errors by a repeated measures ANCOVA. Manual FCP thickness measurements performed by clinical sites were compared against reading center values.

Results: 275 SDOCT volume scans were available for analysis. Scan position was adequate (center of scan within 250 μm radius from FCP) in 203 cases, fair (within 500 μm) in 56 cases and poor (> 500 μm) in 16 cases. The mean absolute deviation after centration of the grid was 11.7±14.9 µm (2.9%) for FCS and 26.4±29.2 µm (8.7%) for FCP. Scan position showed an influence on the measurement error (p< 0.001 for FCS and FCP). Retinal thickness showed an influence on the error for FCS (p< 0.001) but not for FCP (p= 0.320). However, for both FCS and FCP, higher retinal thickness values appeared to be associated with a negative difference between automated and corrected values and vice versa (p< 0.001). Mean absolute difference between FCP values from clinical sites and reading center was 43.7±30.4 µm (15.6%).

Conclusions: Decentration of SDOCT volume scans may result in incorrect retinal thickness measurements in clinical trials for DME even in the absence of segmentation errors. FCP values show higher errors compared to FCS values. Manual assessment by clinical sites did not improve the measurement error. OCT scan position and retinal thickness showed an influence on the measurement error. Manual analysis of retinal thickness by a central reading center is of value in clinical trials for DME.