Abstract
Purpose :
Variability in visual field data in patients with ocular disease is a significant challenge when monitoring for clinical progression or improvements during clinical trials. Omitting points in more severely damaged areas of the field in glaucoma patients reduced test-retest variability of static perimetry and enhanced determination of progression. We tested whether this approach would yield similar findings in patients with retinitis pigmentosa (RP).
Methods :
40 eyes of 20 RP patients, each with visual acuities worse than 20/70, completed the Humphrey 30-2 size V Fastpac test, twice per visit, on 3 separate visits, totaling 6 fields per eye. Threshold values were determined at each of the 76 test points with any values <0dB excluded. The 95% coefficients of repeatability (CR) were calculated for each eye for all threshold values (uncensored), and for those points that only included thresholds above various dB values ranging from 10 to 20 (censored).
Results :
Only 33 eyes in 17 RP patients had test locations with sensitivities ≥0 dB across all six tests. Including all data points (uncensored) resulted in a 95% CR of 8.66 dB across 16.20 test points on average, while censoring points at the lowest level of <10 dB and highest level we explored of <20 dB resulted in 95% CRs of 7.26 dB and 4.91 dB (16% and 43% reduction in variability) for an average of 8.3 and 5.4 test locations for 22 and 7 eyes in 12 and 5 patients, respectively. Censoring at a mid-level of <15 dB resulted in a 95% CR of 6.19 dB (29% reduction in variability) for 18 eyes and 10 patients, which was a trade-off for loss of included eyes and reduction in variability.
Conclusions :
Censoring threshold values of lower sensitivity results in significant reduction in test-retest variability, but has the risk of losing data for eyes with severe visual loss. This approach may be beneficial in monitoring for changes in central visual fields of RP patients, in addition to kinetic perimetry for monitoring the peripheral extent of the field.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.