Abstract
Purpose: :
Test-retest variability (TRV) hinders our capacity to track visual field progression. For example consensus opinion tells us that even if we do 6 fields in 2 years we will only have an 80% chance of detecting a field whose MD is actually progressing at 2 dB/yr[2008 BJO 92: 569-73], pity the younger patient. Undersampling occurs in standard automated perimetry (SAP) if sensitivity varies across a visual field faster than the Nyquist rate (Nq): 1/12 cycles/degree, an unrealistically slow rate of change of sensitivity. Building on our previously published work on undersampling and TRV[IOVS 2011;52:1014-22] the present study sought to quantify the relative contributions of eye movements and undersampling to TRV.
Methods: :
Model visual fields were spatially smoothed in 9 gradations from nil to Nq/4 (very smooth). For each of the 9 levels of smoothing inter-quartile ranges (IQR = 25th to 75th percentiles) of box plots of TRV were determined for 11 bands of scotoma depth from -28.5 to -1.5 dB. This was repeated for 500 sampled fields for each smoothness. Sampling included eye movements equal to that that of good fixation (σ = 0.6°).
Results: :
As reported for SAP fields TRV for the unsmoothed fields grew with scotoma depth, being larger than smooth fields at the 9 largest scotoma depths (all p<0.003). At the 5 test scotoma depths > -28.5 dB the IQRs of those smoothed fields were only 2.3 ± 0.33 dB and so were smaller by 6.0 ± 0.5 dB than the same unsmoothed fields (p<0.0005).
Conclusions: :
Only about 2.3 dB of the IQR of TRV can be attributed to eye movements, the remainder is due to undersampling of the field, which is exacerbated by the small Type III stimulus size. Larger blurry stimuli, which smooth the visual field as it is sampled, should reduce TRV.
Keywords: perimetry • detection • eye movements