Fig 1.
Example of the stimuli used in the handheld radial shape discrimination (hRSD) test[17].
The test employed three radial frequency patterns, presented on an Apple iPod Touch. The vertical position of each pattern was randomly varied, as was the position of the target (distorted) pattern (the pattern on the right in this example).
Table 1.
Study inclusion and exclusion criteria.
Fig 2.
Diagram demonstrating the flow of study participants.
SE: Study eye—no nAMD at recruitment; FE: First eye; diagnosis of nAMD, receiving regular assessment and treatment. CSCR: central serous chorioretinopathy; CNV; choroidal neovascularisation; GA: geographic atrophy; PED: pigment epithelium detachment; VMT: vitreomacular traction; ERM: epiretinal membrane; BRVO: branch retinal vein occlusion; DM: diabetes mellitus.
Fig 3.
Mean (±95% CI) baseline and final VA (A), hRSD (B) results and central subfield thickness (CST; C) for 19 converters (Con) and 160 non-converters (NC).
Note that in C the error bars for the non-converter data are smaller than the symbols.
Fig 4.
hRSD results for the study eyes in the non-converters [SE(NC); final visit], the study eyes in the converters [SE(C); conversion visit] and eyes in which nAMD has been diagnosed tested around the time of the first antiVEGF injection (FE).
Datapoints indicate individual test results, the mean (±95% CI) also shown for each group. The grey horizontal line and region is the mean (±95% CI) hRSD result for healthy eyes, taken from Ku et al (2016). The one-way ANOVA result (comparing the three patient groups) and post-hoc test results (Tamhane’s T2) are shown for these data.
Fig 5.
Timecourse of hRSD performance in the SE of converters (Red) and the SE in the non-converters (Blue).
Solid lines show the Loess fit (±95% CI) calculated by the loess package in R, using a default span parameter of 0.75. The fit was aligned on the conversion or final study visit, working back in time. All available data were used.
Fig 6.
Receiver operating characteristics (ROC) curve for the performance of the hRSD test in distinguishing between converters and non-converters (ie detecting the earliest stage of nAMD).
The area under the curve (AUC; 95% confidence interval) and accompanying p value are shown. Dotted lines indicate a sensitivity of 0.79 and a specificity of 0.54 at an hRSD threshold of -0.60 logMAR.
Table 2.
Effect of varying hRSD cut-offs.