Blumenthal EZ, Sapir-Pichhadze R.
Ophthalmology. 2003 Jan;110(1):196-200.
Department of Ophthalmology, Hadassah University Hospital, and Hebrew University-Hadassah Medical School, Jerusalem, Israel. firstname.lastname@example.org
OBJECTIVE: In this study, the capability of statistical analysis indices to characterize static automated visual fields (VFs) accurately in cases of far-advanced glaucoma was assessed.
DESIGN: Retrospective observational case series.
PARTICIPANTS: Sixteen eyes of 15 patients with end-stage glaucoma and evidence of collapse of VF statistical analysis indices were included in the study.
METHODS: End-stage glaucoma was defined as vertical cup-to-disc ratio of 0.9 or more, mean deviation less than -24 dB and with only a central or temporal island remaining in the VF gray scale. Collapse of statistical indices was defined as any of the following: pattern deviation probability plot without a single VF location showing P < 0.5%; corrected pattern standard deviation (CPSD) and pattern standard deviation (PSD) probability less than 5% or within normal limits (WNL); short-term fluctuation (SF) probability WNL; glaucoma hemifield test (GHT) not outside normal limits (ONL); or presence of a low patient reliability comment triggered by 40% or more false-negative (FN) responses.
MAIN OUTCOME MEASURES: Visual field statistical indices.
RESULTS: Of the 16 VFs showing misleading statistical calculations, 9 of 16 eyes had a normal pattern deviation probability plot. The PSD, SF, and CPSD parameters were normal or barely outside the normal range in 4 of 16, 10 of 16, and 5 of 16 eyes, respectively. The GHT was ONL in 7 of 13 eyes, borderline with generalized reduction of sensitivity (GRS) in three eyes, and only GRS in two additional eyes. Low patient reliability was triggered because of an FN score of 40% or more in 10 of 16 eyes.
CONCLUSIONS: Statistical indices are crucial for the interpretation of automated static VFs. However, in end-stage glaucomatous VF loss, both summary statistical indices and reliability indices may not detect abnormality, thus misleading the casual observer.