Area under the Curve as an Alternative Outcome Measure of Dark Adaptation Response

Abstract

Purpose : Rod intercept time (RIT) is commonly used as a measure of dark adaptation (DA) response. However, in some age-related macular degeneration (AMD) patients, the sensitivity fails to recover to the requisite threshold within the testing duration for a valid RIT determination. This precludes proper quantification of the DA response. We evaluated whether DA response in such eyes can be quantified using the area under the dark adaptation curve (AUDAC), which can serve as an alternative measure of DA response within the same testing time frame.

Methods : DA was measured using the AdaptDx (Maculogix, Harrisburg PA) 20-minute protocol (which output RIT value) in 136 eyes (AMD: 98, Control: 38). The RIT criterion was 3 log units sensitivity recovery, failing which, the RIT was set at maximum of 20 minutes per protocol. Separately, normalized AUDAC was computed from the DA curve using the trapezoid method with respect to the same measurement limits. Association between AUDAC and RIT in eyes with RIT < 20 minutes was computed using a linear model, which was then used to predict the RIT values in eyes that failed to record a valid RIT value. Furthermore, the AUDAC was evaluated for predicting AMD presence using a logistic regression model, and the accuracy and the F score were compared with that for the AdaptDx threshold (RIT=6.5 min.).

Results : In 101 eyes with RIT < 20 minutes, the median [IQR] values of RIT and AUDAC were 5.8 [6.4] minutes and 0.063 [0.054], respectively, with a strong linear association between the two measures (p < 0.001, R² = 0.87). In 35 eyes that failed to record a valid RIT (RIT set at 20 minutes; 34 AMD, 1 control), there was a large variation in the AUDAC (median [IQR]: 0.276 [0.091]) due to the differences in the underlying DA curves, indicating that DA response can still be measured in these eyes. As expected, the RIT was projected to be >20 minutes in 33 out of the 35 eyes (median [IQR] RIT was 24.1 [7.8] minutes) using the linear model between RIT and AUDAC. When predicting AMD presence, the performance of AUDAC (76.5% accuracy; F₁=0.84) was comparable to RIT (78% accuracy; F₁=0.83).

Conclusions : In eyes where sensitivity fails to recover to the requisite criterion, including many AMD patients, the DA response cannot be quantified using RIT. AUDAC offers a reliable alternative analytic approach to measuring DA response for all eyes, even those that fail to record a valid RIT value.

This is a 2021 ARVO Annual Meeting abstract.