Repeatability and reproducibility of manual choroidal thickness measurement using Lenstar images in children before and after orthokeratology treatment

Abstract

PurposeTo investigate the repeatability and reproducibility of choroidal thickness measurements using Lenstar images in young myopic children before and after one-month orthokeratology (ortho-k) treatment. MethodOcular biometry of 39 subjects were performed using the Lenstar 900. The first five measurements with maximum differences of 0.02 mm in axial length in the right eyes were saved and used for measurement of choroidal thickness. Subfoveal choroidal thickness were manually measured by identifying the signals from the retinal pigmented epithelium layer and chorioscleral interface. Repeatability was determined by comparing measurements of the same images made by the same observer on two separate occasions (four weeks apart), while reproducibility was calculated by comparing measurements of the same images made by two independent observers. Data was analysed using intra-class correlation coefficients (ICC) and non-parametric Bland and Altman plots. ResultsThe choroidal peaks could not be identified in all five measurements in all subjects. On average, only 71% subjects had at least four definable images. Compared with the use of fewer than four images, reliability using an average of four definable images improved statistically, but remained clinically unacceptable (>10 µm), although pre- and post-ortho-k ICC values were good to excellent for repeatability (0.867 and 0.975, respectively) and excellent and good for reproducibility (0.959 and 0.868, respectively). Non-parametric pre- and post-ortho-k limits of agreement (2.5% and 97.5% percentiles) obtained were −45.8 to 79.3 µm and −30.3 to 9.5 µm, respectively for repeatability, and −29.0 to 33.5 µm and −21.8 to 70.0 µm, respectively for reproducibility. ConclusionChoroidal thickness measurements using the Lenstar did not show good reliability, despite the high ICC values, non-parametric Bland and Altman plots demonstrated a wide variability of measurement errors. Any changes in subfoveal choroidal thickness, measured by Lenstar, of <80 µm may not represent real changes.