Improving the detection of progression in longitudinal microperimetry data in age-related macular degeneration by using pointwise analysis of data - a MACUSTAR study report

Abstract

Purpose : Single summary (global) measures from microperimetry are commonly used to monitor visual function in trials for age-related macular degeneration (AMD). We evaluate the performance of pointwise (location) analyses in detecting functional progression in intermediate AMD in the MACUSTAR prospective study.

Methods : We used mesopic microperimetry data (MAIA; iCARE, Finland) from 447 study participants with intermediate AMD (Beckman classification by a central reading center) from 20 sites across Europe. Microperimetry (33-point test pattern) was performed at 9 visits over a median of 3 years (IQR 1.5-3.5). We used structural conversion to late AMD observed in 80 participants as the independent reference standard for progression in this study. Two index methods of detecting progression were evaluated: [1] Ordinary rate of change in global mean sensitivity. [2] Trend-based pointwise progression using linear mixed-effects models used in data from standard perimetry in glaucoma. We estimated mean time to detection of progression and proportion of progressors detected (hit-rate) at the same specificity (1 – false positive rate), derived by applying the same methods to random permutations of the test series of 367 eyes which did not demonstrate structural conversion.

Results : There was a greater rate of sensitivity loss in microperimetry measurement in eyes that demonstrated structural conversion compared to those that did not (global mean sensitivity -1.17 (2.5%/97.5% confidence intervals = -0.85/-1.47) vs -0.37 (0.28/-0.46) dB/year, p < 0.0001). Mean time to detection of progression was consistently lower in the pointwise method with a mostly equivalent hit-rate (Figure 1); for example, at a 5% false positive rate the time to progression was 2.76 and 3.09 years for pointwise and global mean sensitivity methods respectively. The pointwise method also had a 2.8 times (CI 1.6/4.3) higher likelihood of detecting progression compared to the global mean sensitivity method (p < 0.0001, Figure 2).

Conclusions : In the analysis of longitudinal microperimetry data, a trend-based pointwise method is quicker at detecting progression compared to using global sensitivity change. The relative gains would likely translate to greater power for trials (shorter duration or fewer participants) where microperimetry is used as a functional endpoint.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.