The correlation with axon counts was strongest for RNFLT (
R = 0.81; 95% CI, 0.73–0.87), compared to either MRW (
R = 0.72; 95% CI, 0.61–0.80) or MRA (
R = 0.70; 95% CI, 0.59–0.79;
P < 0.01 for each comparison;
Fig. 5). This also was true if only EG eyes were considered (
P < 0.05 for each comparison). There was no significant difference, however, between the correlation of MRW to axon count and the correlation of MRA to axon count (
P = 0.43). Comparing across the three panels of
Figure 5 also reveals that there is less overlap between EG and control eyes for MRW and MRA at low levels of axon loss compared to RNFLT. For example, only 43 EG eyes (84%) had a final value of RNFLT below the median value for control eyes, whereas 50 (98%,
P = 0.006) and 51 (100%,
P = 0.001) EG eyes, respectively, had final MRW and MRA values below the control eye median. The EG eyes (red points) tend to shift downward relative to control eyes (blue points) for MRW and MRA, whereas the EG eyes tend to shift leftward relative to control eyes for RNFLT, indicating that axon loss occurs before manifestation of loss by RNFLT. This lag for RNFLT is shown more clearly by comparison of relative loss based on intereye analysis in the next section.
The correlation between relative RNFLT loss and relative axon loss (
R = 0.89) was not significantly stronger than that for MRW (
R = 0.86,
P = 0.17, Steiger's test), but both were significantly stronger than MRA (
R = 0.81;
P = 0.02 each comparison, Steiger's test;
Fig. 6). The data in
Figure 6 also exhibit the phenomenon whereby relative loss of RNFLT is shifted leftward (of the 1:1 line) indicating a “delay” relative to axon loss, whereas the data for MRW and MRA are scattered more symmetrically about the 1:1 line.