Purpose: Visual field trend analysis can be influenced by outlying values that may disproportionately affect estimation of the rate of change. We tested a modified approach to visual field trend analysis to minimize this problem.
Methods: Automated pointwise linear regression (PLR) was used in glaucoma patients with ≥13 SITA-Standard 24-2 VF tests in either eye. In the control group (Group A), conventional PLR using the entire set of VF tests was carried out. In the other 3 groups (study groups), a truncated analysis was done using only the first and last 3 (Group B), first and last 4 (Group C), or first and last 5 (Group D) VF tests. We compared the global slopes (dB/y), number of eyes experiencing significant progression, and significant improvement between groups.
Results: Ninety eyes of 90 patients were evaluated. The mean number±SD of VF tests was 15.7±2.6, spanning 7.8±1.7 years. The study groups showed similar global rates of VF change as the control group (Group A=-0.48±0.5, Group B=-0.48±0.6, Group C=-0.48±0.6, Group D=-0.48±0.5 dB/y, P>0.05), and a similar number of eyes reaching a progression endpoint (Group A=53, Group B=52, Group C=49, Group D=53, P>0.05). However, Group B showed fewer eyes presenting VF improvement (false-positives).
Conclusions: The modified VF trend-analysis showed greater specificity than conventional PLR in a population with glaucoma.