Purpose: To evaluate optical coherence tomography (OCT) reproducibility in patients with diabetic macular edema (DME).
Design: Prospective 1-day observational study.
Participants: Two hundred twelve eyes of 107 patients with DME involving the macular center by clinical examination and OCT central subfield thickness of > or =225 microm.
Methods: Retinal thickness was measured with the OCT3 system, and scans were evaluated by a reading center. Reproducibility of retinal thickness measurements was assessed, and 95% confidence intervals (CIs) for a change in thickness were estimated.
Main outcome measures: Reproducibility of OCT-measured central subfield thickness.
Results: Reproducibility was better for central subfield thickness than for center point thickness (half-width of the 95% CI for absolute change, 38 microm vs. 50 microm, and for relative change, 11% vs. 17%, respectively; P<0.001). The median absolute difference between replicate measurements of the central subfield was 7 microm (2%). Half-widths of the 95% CI for a change in central subfield thickness were 22, 23, 33, and 56 microm for scans with central subfield thicknesses of <200, 200 to <250, 250 to <400, and > or =400 microm, respectively. When expressed as percentage differences between 2 measurements, half-widths of the 95% CI for a change in central subfield thickness were 10%, 10%, 10%, and 13% for scans with central subfield thicknesses of <200, 200 to <250, 250 to <400, and > or =400 microm, respectively. We were unable to identify an effect on reproducibility of central subfield measurements with respect to the presence of cystoid abnormalities, subretinal fluid, vitreomacular traction, or reduced visual acuity. Reproducibility was better when both scans had a standard deviation (SD) of the center point of <10.0% (half-width of the 95% CI for change, 33 microm vs. 56 microm; P<0.001).
Conclusions: Reproducibility is better for central subfield thickness measurements than for center point measurements, and variability is less with retinal thickness when expressed as a percent change than when expressed as an absolute change. A change in central subfield thickness exceeding 11% is likely to be real. Scans with an SD of the center point of > or =10.0% are less reproducible and should be viewed with caution when assessing the validity of an observed change in retinal thickness in patients with DME.