Abstract
Purpose :
For the clinically relevant discrimination between eyes at risk of neovascular age-related macular degeneration (nAMD) and those recently diagnosed with nAMD, we have demonstrated that the hRSD test has moderate sensitivity (Pitrelli Vazquez et al, 2018, PLOS One 13:e0207342). How does the inclusion of different participants (i.e. those with healthy eyes, “high risk” fellow unaffected eyes of patients with unilateral nAMD, eyes with newly diagnosed, untreated nAMD and eyes with established nAMD receiving treatment) modify the diagnostic performance of the hRSD test for the detection of nAMD?
Methods :
In the hRSD test, run on an iPod Touch, participants selected a single distorted circular radial frequency pattern in a 3 or 4 spatial alternative forced choice task. A staircase procedure established the threshold for detecting distortion (Wang et al, 2013, IOVS 54:5497). Visual acuity (VA), and for some participants central subfield thickness (CST), were also measured. In visually healthy participants, one eye was selected at random for analysis; in patients with unilateral nAMD, diagnosed and fellow eyes were tested where possible.
Results :
In 220 healthy participants (mean±SD age 42±15y), hRSD threshold and VA were -0.79±0.12 logMAR and -0.05±0.12 logMAR respectively. For a subset of 50 healthy participants (age:55±14y; hRSD-0.77±0.12 logMAR; VA -0.09±0.11 logMAR), CST was 283±24µm. In 160 high risk eyes (mean age: 79±8y), hRSD and VA were -0.58±0.19 logMAR and 0.05±0.12 logMAR respectively; CST was 276±28µm. In 119 eyes with diagnosed nAMD there was a further decline in function (hRSD: -0.19±0.34 logMAR; VA: 0.31±0.24 logMAR). In an omnibus ROC analysis incorporating all these data (N=499), hRSD performed slightly better than VA at identifying eyes with nAMD (AUC±SE; hRSD: 0.94±0.1; VA: 0.89±0.2; both p<0.001). For the optimum hRSD cut-off value (-0.57 logMAR), sensitivity was 0.92 and specificity 0.80. In an ROC analysis of 33 healthy participants aged >50y, the 160 high risk eyes and 19 eyes with newly diagnosed but untreated nAMD, test performance dropped, but AUC remained greater for hRSD than for VA (0.74±0.06, p=0.001 vs 0.68±0.07, p=0.014); both were outperformed by CST (0.92±0.03, p<0.001).
Conclusions :
The hRSD test outperformed VA in detecting nAMD. When comparing the hRSD and other tests, careful consideration should be given to the participant composition of studies.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.