Purchase this article with an account.
J. W. Pak, T. Pauli, C. Chandler, E. Treichel, S. Reed, A. Domalpally, R. P. Danis, Jr., L. D. Hubbard; Lens Opacity Grading Performed Independently vs. Longitudinally. Invest. Ophthalmol. Vis. Sci. 2008;49(13):1924. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Independent evaluation of lens opacities from photographs can introduce biologically implausible results over time (posterior subcapsular cataract [PSC] "regression" from presence to absence). We explored an alternative approach, which is comparative and unmasked to earlier images and grades, to minimize "noise" from regression.
From various clinical studies at FPRC having film fundus reflex images (in effect, retro-illumination), we selected a sample of 48 eyes having a wide range of PSC (0 - 100%), 3 visits each (Baseline, Month 4 and Year 1), and varying photo quality. PSC in the 5 mm central zone was evaluated according to the AREDS classification (Report #4, AJO 2001). Four evaluators each utilized two different approaches: independent grading (IG - done masked to other visits) and comparative or longitudinal grading (LG - done with access to earlier images and grades, but without ability to change the latter). First, the IG evaluation was done, masked to subject and visit, with images from the same subject several days apart to minimize recall. Then, LG evaluation was performed, one week after the IG evaluation, again to reduce recall.
(1) Among 48 eyes, PSC regression was observed with IG in 29.2% over all three visits and 17.7% over only two visits (baseline/year 1). LG more than halved PSC regression, to 13.0% over three visits and to 7.3% over two visits. (2) Within evaluator correlation coefficients between IG and LG are high, both at baseline (mean R2 = 0.84 [SD ±0.11]) and at year 1 (mean R2 = 0.92 [±0.07]) showing that IG and LG produce similar overall results. (3) However, between evaluator comparisons (relevant because IG may use different graders at different visits) show lower correlations regardless of mode, especially at baseline (IG mean R2 = 0.34 [SD ±0.25]; LG mean R2 = 0.46 [SD ±0.24]) but also at year 1 (IG mean R2 = 0.78 [SD ±0.17]; LG mean R2 = 0.72 [SD ±0.14]).
In our study, LG reduced PSC regression by more than half compared to IG. Highest reproducibility for PSC was achieved by adopting LG, using the same evaluator for baseline and follow-up visits. While IG may reduce bias from the belief that cataract always worsens over time, the methodological "noise" in IG for PSC (due to uneven image quality and subjectivity of evaluation - both of which LG takes into account) increases biologically implausible cataract "regression."
This PDF is available to Subscribers Only