To determine how well rates of localized retinal nerve fiber layer thickness (RNFLT) change correlate with rates of sensitivity change at corresponding locations in the visual field in glaucoma.
Retrospective cohort study.
Three hundred and sixty-four eyes of 191 participants with suspected or confirmed glaucoma, as judged by experienced clinicians, were tested every 6 months with perimetry and optical coherence tomography (OCT). For each 24-2 visual field location, the corresponding sectoral peripapillary RNFLT was defined using a 30-degree sector, centered on the angle of nerve fiber entry into the optic nerve head. Rates of change of pointwise sensitivity and sectoral RNFLT were calculated over the last 8 visits at which reliable data were obtained. Passing-Bablok regression was used to predict the rate of pointwise sensitivity change from the rate of sectoral RNFLT change, for each location.
Rates of sectoral RNFLT change were significantly predictive of rates of pointwise sensitivity change at all locations in the field. Correlations were modest, averaging 0.15, ranging from 0.03 to 0.25 depending on the location. A 1 μm/y more rapid thinning in corresponding sectors was associated with 0.3 dB/y more rapid loss in the superior visual field but less than 0.1 dB/y more rapid loss at many locations in the inferior visual field.
Localized RNFL thinning is associated with sensitivity loss at corresponding locations in the visual field, and their rates of change are significantly correlated. Peripapillary RNFLT may be used to monitor localized changes caused by glaucoma that have measurable consequences for a patient’s vision.
Optical coherence tomography (OCT) allows visualization and measurement of ocular structures with high resolution in 3 dimensions. Glaucomatous eyes have been shown to have reduced retinal nerve fiber layer thickness (RNFLT), optic nerve head minimum rim width (MRW), and macular inner retinal layer thickness, among other structural abnormalities that can be quantified using OCT. Furthermore, reductions in these measures correlate with reduced function, and imaging tests are preferred over perimetry by patients. Therefore it has been suggested that structural measures derived from OCT could be used both diagnostically and in clinical trials. Since preserving function is the ultimate goal of almost all interventions in glaucoma, it is essential to know how closely OCT relates to commonly used functional measures.
Unfortunately, correlations between structural and functional measures are not as strong as may be anticipated. We have previously reported that the correlation between global average peripapillary RNFLT and mean deviation (MD) from standard automated perimetry was modest, with r = 0.646; and that the correlation between MRW and MD was weaker still ( r = 0.546). The strength of these correlation coefficients depends on the range of severities of glaucomatous damage in the cohort, but is similar to studies in other cohorts (eg, r = 0.549 ). When localized to smaller regions of the retina and associated visual field, reported correlations range from as high as 0.81 down to around 0.3, depending on both the cohort and the size of the regions studied. Owing to this weakness of the structure-function relation, concerns have been raised by the Food and Drug Administration in the United States about the sole use of structural measures of progression as endpoints in clinical trials.
There are at least 3 plausible explanations for this dissociation. It is possible that structural and functional changes, as measured clinically, do not coincide; it has been suggested that structural changes may precede functional changes along a continuum, with important clinical consequences and mechanistic implications. Alternatively, it may be that the substantial test-retest variability observed in clinical test results obscures the true strength of the underlying structure-function relation. Using confocal scanning laser tomography, we showed that the test-retest variability was sufficiently high that it could be the sole cause of the structure-function dissociation. In addition, we have previously shown that the intertest variability (defined as the standard deviation of residuals from the longitudinal trend) was as high as 170% of the typical (median) annual rate of change in RNFLT, and 230% of the median rate of change for MRW, among managed glaucoma patients ; and that perimetry was as much as 80% more variable than RNFLT when compared to their respective rates of change. A third possibility is that each individual’s visual system is tailored to the number of retinal ganglion cells present in that eye, for example by adjusting gains associated with retino-cortical connections. There is considerable interindividual variability in RNFLT for the same level of sensitivity in healthy eyes, and this could contribute to the structure-function dissociation observed in cross-sectional studies of glaucomatous eyes.
Although studies that examine the cross-sectional correlation between OCT and perimetry have been performed, the strength of the correlation between the longitudinal rates of change in these measures remains uncertain. As noted above, there is a considerable range of RNFLT for normal eyes with the same level of function. It is therefore possible that structural and functional measures could progress in perfect correspondence for each eye separately, while still resulting in a weak cross-sectional structure-function relation owing to this population variability. This would allow structure and function to be used interchangeably for monitoring progression, provided that a reliable baseline has been determined for each.
Evidence is now emerging that there is indeed a significant correlation between rates of structural change (from OCT) and functional change (from automated perimetry). We have previously reported a significant correlation of r = 0.361 between the rates of change of MD and global average RNFLT. However, glaucomatous disease manifests in a predominantly localized or asymmetric manner, both between the 2 eyes of an individual and within an eye. To our knowledge, no studies have yet looked at the strength of the correlation between the rates of localized functional and structural change using these testing modalities. The use of global measures of progression such as MD reduces variability (in the same way that the mean of 2 measures will be less variable than either one alone), and so it remains possible that for pointwise or localized analyses, the structure-function correlation between rates of change may not be statistically significant; and/or it may be that the correlations observed when global rates of change are examined are an artifact of noncorresponding regions of the visual field and RNFL changing at the same time with a temporal disconnect. In this study, we ask whether the rates of localized RNFL thinning correlate with the rates of sensitivity reduction at corresponding areas of the visual field. Such a relation would support the principle of using OCT to supplement functional assessment via automated perimetry in clinical applications.
Data for this study were obtained from participants in the Portland Progression Project (P3), a prospective longitudinal cohort study of the course and risk factors for glaucomatous progression. Participants were recruited to and tested at a tertiary glaucoma clinic at the Devers Eye Institute in Portland, Oregon, USA. Inclusion criteria were a diagnosis of primary open-angle glaucoma and/or likelihood of developing glaucomatous damage (eg, ocular hypertension with other risk factors such as a suspicious-looking optic disc or a family history of glaucoma), as determined by each participant’s physician. A visual field defect was not a requirement for study entry, nor was having an abnormal RNFLT. Exclusion criteria were an inability to perform reliable visual field testing, best-corrected visual acuity worse than 20/40, cataract or media opacities likely to significantly increase light scatter, or other conditions or medications that may affect the visual field. Participants provided written informed consent once all of the risks and benefits of participation were explained to them. All protocols were approved and monitored by the Legacy Health Institutional Review Board and adhered to the Health Insurance Portability and Accountability Act of 1996 and the tenets of the Declaration of Helsinki.
Participants were tested using a variety of methods, including automated perimetry and OCT, approximately every 6 months. Since damage may not occur at a constant rate over long periods of time owing to variation in pathophysiological course and/or treatment decisions, data were used from the most recent 8 visits at which reliable measurements (as outlined below) were acquired.
Functional testing was performed using automated white-on-white perimetry with a Humphrey field analyzer (HFA II; Carl-Zeiss Meditec, Dublin, California, USA), with a size III stimulus, SITA standard algorithm, and 24-2 test pattern. Visual field tests were excluded from analysis if they had greater than 20% false positives or 33% fixation losses. For the primary analysis, pointwise sensitivities in decibels (dB) were used, as output by the perimeter; these will be denoted Sens dB . Because it has been reported that the structure-function relation may be nonlinear, and that functional progression may also be nonlinear when expressed in dB, a secondary analysis was performed expressing pointwise sensitivities on a linear scale of 10 (dB−30)/10 , such that 0 dB→0.001, 10 dB→0.01, 20 dB→0.1, 30 dB→1.0, etc. Sensitivities on this scale, which will be denoted Sens Lin , may then be more linearly related to RNFLT, and hence to axon counts; and may also progress more linearly over time. A third analysis was performed after setting all pointwise sensitivities below 15 dB to equal exactly 15 dB, on the basis that sensitivities below 15–19 dB are unreliable ; this has been shown to aid the detection of pointwise change in glaucoma. Pointwise sensitivities that have undergone this process will be denoted Sens 15 .
Spectral-domain OCT was performed using a Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany). The peripapillary RNFLTs were derived on each test date from a single circular B-scan with a radius of 6 degrees centered on the optic disc. The operator centered the position of the circular B-scan on the optic disc at baseline, and each follow-up scan was registered in real time to the location of the baseline reference scan for each eye using the instrument’s automated eye-tracking capability. Nine to 16 sweeps were averaged to comprise the final recorded B-scan for each eye on each test date. Scans with quality score less than 15 were excluded, as recommended by the manufacturer. The instrument’s automated segmentations of the inner limiting membrane and the posterior border of the RNFL were manually adjusted, if necessary, by experienced technicians in order to address obvious segmentation errors (without reference to the functional results or knowledge of this study’s specific hypothesis).
The “average angles” where nerve fiber bundles enter the optic disc from each of the 52 non–blind spot visual field locations were taken from the map of Garway-Heath and associates. Because the exact angle of entry will vary between eyes, and may vary between scans for the same eye based on factors such as head position during scanning, sectoral RNFL thicknesses were calculated within peripapillary sectors centered on each of these average angles, rather than just using the value from a single A-scan exactly corresponding with the average angle. In order to smooth out artifacts caused by blood vessels or other irregularities being just inside or just outside the sector edge, a trapezoidally weighted average thickness was used. A-scans within ±10 degrees of the average angle were assigned a weighting of 1.0. A-scans within 10–20 degrees either side of the average angle were assigned weightings that decreased linearly with distance, from a weighting of 1.0 at ±10 degrees down to a weighting of 0.0 at ±20 degrees from the average angle. Thus these sectors are conceptually equivalent to a 30-degree-wide sector centered on the average angle of entry, and they will henceforth be referred to as “30° sectors,” and the sectoral thickness will be denoted RNFLT sector .
Analyses were performed using the R language and environment for statistical computing (Version 2.15.3; R Core Team, Vienna, Austria; 2013, www.R-project.org/ ). For each non–blind spot visual field location in each eye, the rates of change of Sens dB , Sens Lin , Sens 15 , and RNFLT sector for the corresponding sector were calculated over the most recent 8 visits by least-squares linear regression.
In order to determine whether the rate of localized structural change is a statistically significant predictor of the rate of pointwise functional change, global analyses were performed. Mixed-effects models were formed to predict the rate of pointwise functional change (either d / dt Sens dB , d / dt Sens Lin , or d / dt Sens 15 ), using the concurrent rate of localized structural change d / dt RNFLT sector , at all 52 non–blind spot visual field locations concurrently. The models were of the form:
d dt Sens = Int + ε ID + ε Eye + ( λ + λ Loc ) d dt RNFL T sector