Monitoring glaucoma is vital, but a new study suggests we shouldn’t rely exclusively on changes in the circumpapillary retinal nerve fiber layer (cpRNFL) thickness to detect progression.

The study assessed the effects of local defects, segmentation errors and improper image alignment on this OCT metric, a commonly used summary statistic that was developed to scale down the amount of information available and reduce variability.

The study included 150 eyes with suspected or early glaucoma. Based on baseline and follow-up OCT imaging, researchers identified 30 eyes as “likely progressed” and 61 eyes as “likely not progressed.” This outcome supports previous findings that suggest the cpRNFL thickness metric performs poorly, with no cutoff that results in high sensitivity and specificity for detecting glaucoma progression.

“When an eye shows a loss of 5μm or more in change in global average thickness, this is considered quantitative evidence of glaucoma progression,” the researchers wrote. “However, recent work suggests that this ‘rule of five’ has poor specificity, i.e., it leads to too many false positives over time.” In fact, the team found that the criterion of 4μm was the most accurate at 77%, with five false positives and 16 false negatives.

According to the investigators, segmentation errors and local defects were responsible for the majority of the false readings; however, there were also segmentation errors in true positive and true negative results. This evidence goes hand-in-hand with previous studies that reported that the cpRNFL thickness metric is affected by segmentation and alignment errors and can miss local defects.

“Global measures, such as changes in average cpRNFL thickness, will miss progression of glaucoma,” the study authors concluded. “As it is an average, it can also miss local defects, or in other words be more likely to detect widespread thinning.” They noted that a more robust approach is necessary for measuring progression and that clinicians should focus more on the agreement among the B-scans, thickness map and probability and deviation plots.