Abstract
Purpose :
To investigate global longitudinal structure-function (SF) relationships between macular ganglion cell complex (GCC) and central visual field (VF) mean deviation (MD) rates of change with a novel Bayesian joint bivariate longitudinal model.
Methods :
117 eyes (117 patients) with 2 years of follow-up and 4 visits were enrolled. Global macular GCC measurements were estimated by averaging the GCC values across 49 superpixels from the Posterior Pole Algorithm map (Spectralis OCT). We fit a Bayesian joint bivariate model with random intercepts (estimated baseline thickness), slopes (rates of change), and residual standard deviations (SD) to investigate the longitudinal SF relationships between global GCC and 10-2 VF MD rates of change. Main outcome measures were the correlations between GCC and VF MD intercepts, slopes, and residual SDs. The proportion of significantly negative and positive slopes for global GCC and VF MD were reported. We considered correlations and rates of change significantly negative or positive when the upper or lower limit of the 95% credible interval (CrI) was less than or higher than 0, respectively.
Results :
Average (SD) baseline MD and follow-up time were –8.3 (5.2) dB and 5.0 (0.9) years, respectively. The mean (95% CrI) global intercept, slope, and residual SD for GCC were 73.2 (71.4, 74.9) µm, –0.42 (–0.52, –0.31) µm/year, and 0.99 (0.88, 1.12) µm. The corresponding numbers for MD were –8.7 (–9.7, –7.7) dB, –0.24 (–0.34, –0.13) dB/year, and 1.21 (1.10, 1.33) dB, respectively. The mean (95% CrI) correlation was 0.47 (0.32, 0.60) between GCC and MD intercepts, 0.23 (0.00, 0.44) between GCC and MD slopes, and 0.16 (–0.08, 0.39) between GCC and MD residual SDs. The proportion of significantly negative slopes for GCC and MD were 50.4% and 20.6% (McNemar’s test: p<0.001); the corresponding proportion of positive slopes were 0.9% and 2.6% (p=0.32).
Conclusions :
We present the first instance of a joint model exploring simultaneous changes of macular structural and functional measures. The highest SF correlations were between cross-sectional measures. Correlation of global structural and functional rates of change was weak over clinically relevant follow-up periods. This model showed a higher likelihood of detecting structural than functional change. Our joint global bivariate model serves as an initial step towards designing local longitudinal SF models.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.