The second objective was to quantify how eye movements contributed to errors in the projection of perimetric stimuli on the retina. Here, we used fixation and projection data from the CMP and structural data from a spectral-domain OCT to model what would happen in conventional perimetry. We specifically isolated the effect of gaze attraction from projected stimuli in what we called evoked displacements. We found that removing these evoked displacements from fixation data significantly shrank the 95% BCEA in both healthy subjects and glaucoma patients (
P < 0.001). This reduction was significantly more pronounced in healthy subjects (
P = 0.034), as evoked displacements were significantly more frequent in this group (
P = 0.047). This could be partially explained by the fact that healthy subjects were less experienced with perimetry than glaucoma patients; however, rather than an actual change in fixation behavior, we attribute this difference to a higher number of seen presentations in healthy subjects, resulting from the way threshold strategies probe VF sensitivity.
5,33 This is also corroborated by the finding that the 10-2 MD was positively correlated with the 95% BCEA in glaucoma subjects (larger for more initial damage), but such a relationship was not significant when evoked displacements were removed. Notably, all glaucoma subjects were experienced test takers. We then found that the projection error of stimulus presentations preceded by an evoked displacement was significantly increased compared to the rest of the presentations (
P < 0.001) (
Fig. 4). This is not an obvious result, as the time interval between presentations (
Table 2) could allow subjects to return to central fixation. Previous work investigating fixation area in fundus perimetry found a significantly increased 95% BCEA in glaucoma patients.
30,31 In a previous study,
23 however, we analyzed data from the PRL assessment phase in the CMP on a different dataset and found no difference between healthy subjects and glaucoma patients, irrespective of their level of damage, although there was a significant difference in other fixation metrics.
23 This is confirmed by the results of this study, as no difference was found in the 95% BCEA between glaucoma and healthy subjects. Interestingly, Longhin et al.
30 reported an increase in BCEA during the perimetric test compared to the initial PRL assessment phase, during which time no stimuli were projected. They speculated that this spread in fixation area could be the effect of projected stimuli attracting fixation, and this is consistent with our findings.