Abstract
Purpose:
To determine the influence of visual field (VF) test parameters on VF reliability.
Methods:
Reliability was determined for each VF in eyes with at least 5 VF tests. Reliability was judged by: (1) the difference (residual) between the measured (actual) and predicted VF MD, and (2) the presence of a residual MD more than 2 standard deviations outside the norm for the studied eye. Predicted VF MD was calculated in models incorporating baseline MD, MD of each study period VF, time from baseline VF date, and the number of annual VFs. Studied predictors of VF reliability included percentage of false negatives, false positives and fixation losses, test duration, time of VF testing, day of VF testing, and season of year.
Results:
1,778 eyes in 1,047 patients were examined, and mean duration between the first and last VFs was 5.1 years. False positives were associated with a greater measured MD than predicted (+1.05 dB/10% false positives, p<0.001) while false negatives were associated with a lower measured MD than predicted (-0.40 dB/10% false negatives, p<0.001). Fixations losses were not associated with residual MD (+0.02 dB/10% fixation losses, p=0.07), while longer test duration was associated with a lower measured MD than predicted (-0.34 dB/extra minute of testing, p<0.001). Residual MD also increased in winter as compared to non-winter months (p≤0.02), but did not vary by time of day or day of week (p>0.05). The odds of a VF demonstrating a residual MD at least two standard deviations outside the norm for the studied eye increased with greater test duration (Odds Ratio [OR]=2.07, p<0.001), false positive percentage (OR=1.26/10% increase, p=0.004), and false negative percentage (OR=1.51/10% increase, p<0.001), but was not associated with the percentage of fixation losses, time of day, day of week, or season of year (p>0.1 for all).
Conclusions:
False positives, false negatives and test duration are the most significant predictors of VF reliability, and should be the primary clinical criteria for determining whether or not a VF result can be considered reliable. Fixation losses do not significantly predict VF reliability independent of other factors and may not have significant meaning in determining whether a VF result can be trusted. False positives are associated with the greatest difference between measured and predicted VF MD, and even modest amounts of false positives should be recognized as potentially producing significant errors.
Keywords: 758 visual fields