Purchase this article with an account.
Nariman Nassiri, Simon Law, Anne Coleman, Joseph Caprioli, Kouros Nouri-Mahdavi; Weighting of Visual Field Mean Deviation according to Test-Retest Variability of Pointwise Thresholds. Invest. Ophthalmol. Vis. Sci. 2013;54(15):3965. doi: https://doi.org/.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To determine whether a ‘corrected’ mean deviation (MD) index weighted according to test-retest variability of the visual field (VF) test locations at baseline, regardless of topographic location, would improve detection of glaucoma progression or prediction by reducing longitudinal noise in visual field series.
274 glaucoma eyes (207 patients) with ≥6 SITA-standard VF exams over at least 3 years of follow-up were reviewed. Out of the above cohort, 84 eyes with three or more locations marked as “X” on GPA follow-up printouts were selected. ‘Corrected’ MD for each visual field exam was calculated weighting the threshold at each test location based on the test-retest variability of the baseline threshold as reported by Heijl et al. (AJO 1989). Linear regression analysis (LRA) of MD and ‘corrected’ MD against time was carried out for the entire follow-up period and the root mean square (RMS) of residuals from the MD and ‘corrected’ MD fits were compared. For comparison of predictive performance, 33 eyes with at least 10 VF exams and 6 years of follow-up were selected and LRA was performed on the first half of data and the observed-minus-predicted values at the end of follow-up were compared.
The 84 study eyes were followed for 8.5±2.4 years and had an average of 10.8±2.8 VF exams. The average (±SD) MD was −4.2 (±4.5) dB at baseline. The median (IQR) decay rate was −0.16 (−0.42 to 0.06) dB/year for MD and −0.14 (−0.37 to 0.05) dB/year for 'corrected' MD (p<0.007). LRA of MD (vs. ‘corrected’ MD) against time detected worsening (i.e. decay rate <0 dB with p<0.05) in 26 (31%) vs. 19 (23%) eyes, and improvement (decay rate>0 dB and p<0.05) in 1 (1%) vs. 4 (5%) eyes, respectively (p<0.001). The median RMS of residuals was 1.06 and 1.04 dB for the MD and ‘corrected’ MD, respectively (p=0.03). The difference between observed and predicted MD (median 1.3, IQR: −0.8 to 5.1 dB) was smaller than the difference between observed and predicted ‘corrected’ MD (median: −4.8, IQR: −6.3 to −1.4 dB)(p=0.027 for the difference in the absolute values).
Weighting of the global index MD according to test-retest variability of VF locations at baseline did not improve detection of glaucoma progression or its prediction. Other weighting strategies need to be explored especially in cohorts with advanced glaucoma where significant long-term variability is expected.
This PDF is available to Subscribers Only