Purpose:
To compare the abilities of PROGRESSOR and GPA2 algorithms to detect glaucomatous VF progression.
Methods:
We reviewed the charts of consecutive glaucoma patients seen in a 2-month period with repeatable VF loss in the first two reliable VF exams of a series of ≥8 SITA-Standard 24-2 VFs, and no other ocular conditions causing VF loss. VF progression was evaluated using 3 methods: A) PROGRESSOR - automated pointwise linear regression (Medisoft, Inc); B) GPA2 visual field index (VFI) regression; and C) GPA2 EMGT criteria (Carl Zeiss Meditech, Inc). For PROGRESSOR, progression was defined if the eye developed at least 2 adjacent test points progressing >1.0 dB/yr, p<0.01. For GPA-VFI, progression was defined if the rate of VFI exceeded 1.0%/yr, p<0.01. The location of progressing points was compared using the Garway-Heath map sectors. The same sequence of VF testing, including baseline and follow-up examinations, was applied to the 2 methods.
Results:
130 eyes (130 patients; mean age 63.6±12.7 yr; mean number of VF, 11.2±3.6; mean follow-up time, 7.3±1.8 yr) were entered. There was a significant correlation between VFI and PROGRESSOR global rates of change (r=0.81, p<0.01). The number of eyes progressing based on PROGRESSOR, GPA-VFI, and the EMGT criteria were 55, 37, and 46, respectively (p=0.06). There was a moderate agreement between PROGRESSOR and the EMGT criteria (k=0.43). There was a moderate agreement between PROGRESSOR and GPA-VFI (k=0.57) and between GPA-VFI and EMGT (k=0.45). PROGRESSOR tended to detect more progressing points per eye than GPA2 (3.7±5.8 vs. 3.1±4.4, p=0.50). When both PROGRESSOR and EMGT criteria detected progression, 94% of the progressing locations occurred within the same VF sector. PROGRESSOR detected more progressing eyes than EMGT regardless of follow-up time, the difference being greater in the quartile with the longest follow-up time (median,9 yrs; 18 vs. 11, p=0.06).
Conclusions:
The two methods showed statistically similar sensitivities for detection of VF progression with moderate agreement for detection rates and excellent agreement for topographic correlations. PROGRESSOR tended to be more sensitive than EMGT in detecting progression which was independent of follow-up time. The absence of a gold-standard to define progression prevents any conclusion regarding the specificity of these methods.
Keywords: visual fields • detection