Abstract
Purpose :
The aim is to test the feasibility of measuring a latent disease state variable for Retinitis Pigmentosa (RP) that can be used to estimate the rate of RP progression and provide a single outcome measure, based on a battery of clinical observations, for emerging treatment trials.
Methods :
Measures of visual acuity, II-4e and V-4e Goldmann visual field areas, long and short wavelength dark-adapted thresholds at fixation, and contrast sensitivity from a previous IRB-approved natural history study of 1087 patients diagnosed with typical RP were compiled over repeated annual visits for a grand total of 4774 patient-visit data sets. Each of the 6 different psychophysical measures (items) were binned into 4 to 7 ordinal categories, for which lower ordinal values signified less severe disease. Rasch analysis was performed on the data matrix using the Masters partial credit model. All measures were expressed as logits (log-odds), which define units of measurement for RP severity on an interval scale.
Results :
Figure 1 illustrates the relative frequency distribution of estimated RP severity measures for the 4774 patient-visits (dotted curve), which is well-centered on the distribution of interval boundaries of the ordinal bins for each item (color-coded hash marks on the abscissa). Figure 2 shows that estimated item measures are lowest (most sensitive to RP severity) for dark-adapted foveal thresholds and highest (least sensitive) for visual acuity. The estimated person and item measures explain 76% of the observed variance. Infit mean squares are distributed as expected when the latent variable has a single source of variance plus random error. Estimated person measures decrease linearly with time at an average rate of 0.2 logit per year (SD=0.34).
Conclusions :
The results are consistent with the existence of a single disease state variable on a continuous severity scale for RP. This study establishes the feasibility of further developing and refining the scale for various types of RP by including additional clinical indicator measures such as ERG, static perimetry, and fundus gradings.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.