Purchase this article with an account.
A. R. Glassman, for the Diabetic Retinopathy Clinical Research Network; Comparison of Optical Coherence Tomography (OCT) Retinal Thickness (RT) Measurements in Diabetic Macular Edema (DME) With and Without Reading Center (RC) Manual OCT Grading. Invest. Ophthalmol. Vis. Sci. 2008;49(13):3485.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To analyze the value of RC correction of errors in automated RT measurements in eyes with DME.
6,522 Zeiss Stratus OCT3 scans obtained in 7 DRCR.net studies were included in the analysis. The RC evaluated whether the automated center point (CP) measurement appeared correct, and when not, it was manually measured using calipers. Four factors CP standard deviation as a percentage of thickness (SD), CP thickness, signal strength, and analysis confidence were evaluated by logistic regression for their association with an automated measurement error (manual measurement needed and exceeded 12% of automated thickness). Curves were constructed to further evaluate each factor by plotting the error rate in the data against the proportion of scans sent to the RC. Using data from one DRCR.net trial, analysis using only automated values was compared with one in which errors were corrected. The impact of measurement error on statistical power was also assessed.
CP SD was the best predictor of an automated measurement error. Based on SD, an error rate of 5% or less was achieved by sending only 33% of scans to the RC (those with a SD ≥5%, Figure 1). Analysis of the data from a completed randomized trial showed that correcting automated errors had no appreciable effect on the results and had little impact on a trial’s statistical power.
In DME clinical trials, the error involved with using automated OCT CP measurements is sufficiently small that results are not likely to be affected if scans are not routinely sent to a RC, provided adequate quality control measures are in place. When needed, a greater degree of accuracy requires sending only about 1/3 of scans to the RC for assessment.
This PDF is available to Subscribers Only