The Effect of Testing Reliability on Visual Field Sensitivity in Normal Eyes: The Singapore Chinese Eye Study

Abstract

To quantitatively determine how the reliability indices in standard automated perimetry (SAP) affect the global indices of visual field (VF) results in nonglaucomatous eyes. Observational, cross-sectional study. A total of 830 adults aged 40 to 80 years, without visual impairment, glaucoma, significant cataract, and major eye diseases, were selected from the population-based Singapore Chinese Eye Study (SCES). Study participants underwent a comprehensive and standardized ocular examination and VF assessment using a Humphrey Field Analyzer II (Carl Zeiss Meditec, Inc., Dublin, CA). The effects of the test reliability, as indicated by the false-negative (FN), false-positive (FP), and fixation loss (FL) rates, on global indices, as indicated by the mean deviation (MD) and pattern standard deviation (PSD), were analyzed with multivariable regression models. The MD and PSD. A total of 1828 VF results from 1235 normal eyes of 830 study subjects were included in the analyses. The multivariable regression analyses adjusted for age, gender, best-corrected visual acuity, and test duration showed that at lower frequencies of false answers (<15%), FNs decreased the MD (β [change in decibels {dB} per 5% increment in false answers]=-0.71 dB; P < 0.001), whereas FPs increased the MD (β= 0.65 dB; P< 0.001). At higher frequencies (≥15%), the false answers influenced the MD to a greater extent, where the β for the associations with FN and FP rates was-1.15 and 1.26 dB, respectively (both P < 0.001). We also found that when FN rate was <15%, higher FN rate increased the PSD (β= 0.51 dB; P < 0.001), and the effect was slightly larger when FN rate was ≥15% (β= 0.71 dB; P < 0.001). The effect of FPs on PSD was observed only when FP rate was <15% (β=-0.22 dB; P < 0.001). The FL had no associations with the MD, and had minimal effects on the PSD. We quantified the effect of unreliable responses on the MD and PSD in SAP. Our study mayallow clinicians to estimate how VF results are affected by varying degrees of unreliability, instead of relying on cutoff values for reliability indices.