Primary primary open-angle glaucoma (POAG) can be considered as a member of the family of diseases that exhibit an optic neuropathy which is characterised by progressive neuro-retinal rim thinning and evacuation, manifested as a loss of the retinal ganglion cells in the inner retina and their axons in the retinal nerve fibre layer (RNFL) and a deformation of connective tissues supporting the optic nerve head, together with a corresponding visual field defect.1-2
While the definition of POAG has not materially changed in recent years, the sophistication of the investigative techniques used for the detection and management of glaucoma is currently moving at an unprecedented pace with the advent of in-vivo high resolution imaging using spectral domain (and other modes of) optical coherence tomography (OCT). Such advances are not only resulting in the detection of glaucoma at an earlier stage but are also aiding the understanding of the disease process, itself. In parallel with the developments in OCT, is the re-alignment of the importance of the visual field examination using standard automated perimetry.3-5
It should be stressed that the definition of POAG includes measures of both structure and function, ie, both assume equal importance in the detection and assessment of the disease. In the early stages of POAG, abnormalities of structure and function can occur concurrently in some patients, while other patients seemingly manifest detectable abnormality of one modality prior to the other. In addition, there can be considerable variation between the structural appearance and the functional outcome across individuals.
The advantage of high resolution imaging, together with the associated quantitative capability, is obvious in the characterisation of the optic nerve head and of the RNFL but such techniques are also enabling an evaluation of the retinal ganglion cell layer and of the lamina cribrosa.
This article will review the ‘best/preferred practice’ approach to the detection of POAG with particular emphasis on the evaluation of structural and functional integrity. The interpretation of the printout will then be discussed. This will cover spectral domain OCT of the optic nerve head, of the peripapillary RNFL and of the retinal ganglion cell layer/inner plexiform layer complex. In Part 2 standard automated perimetry will be covered. The utility of a combined structural and functional approach to the examination of patients with glaucoma will be illustrated.
The characteristics of the optic nerve head associated with POAG are long established and include an increase in cup size, an increase in cup to disc ratio, disc asymmetry, changes in the lamina cribrosa, loss of neuroretinal rim, pallor, increase in peripapillary atrophy, vessel changes and disc margin haemorrhage6 (Figure 1).
[CaptionComponent="2742"]
Such appearances are best viewed stereoscopically and both the qualitative and quantitative documentation depend upon the skill of the clinician. The associated abnormality of the RNFL has traditionally received less attention but can be documented by red-free viewing (Figure 2).
[CaptionComponent="2743"]
Spectral domain optical coherence tomography
Optical coherence tomography is an objective non-invasive in-vivo three-dimensional imaging technique which uses the principle of low-coherence interferometry to image the light backscattered from a given structure. For ocular investigation, the light source is a broadband superluminescent diode, generally centred at 840nm. Light from the diode is split into two optical beams. The sample beam is scanned across the given structure and the reference beam travels to a mirror. The returning beams are recombined at a detector which, in the case of spectral domain (also known as Fourier domain) OCT, is a spectrometer. The relative amplitudes and phases of the spectral components of the recombined light returning from throughout the depth of the structure under investigation are then processed using Fourier analysis. The required sequence of one-dimensional depth scanning (A-scans) form a series of B-scans that generate a three-dimensional image of the given structure. Current generation clinical spectral domain OCT systems can acquire approximately 25,000 to 40,000 A-scans per second. The axial resolution of spectral domain OCT is between 5 and 10µm.
OCT was initially applied in eye care for the investigation of vitreo-retinal and macular disease. However, the technique has now also radically altered the approach/emphasis to the detection and management of glaucoma. It enables quantitative measures not only of the optic nerve head but also of the peripapillary RNFL thickness and of the combined macular ganglion cell/inner plexiform layer thickness. These measures are then compared to age- and/or disc size- corrected normal values. The assessment of the macular ganglion cell/inner plexiform layer thickness in glaucoma is currently the subject of much research interest.7-8
For imaging the optic nerve head with the Cirrus HD-OCT (Figure 3), for example, a 6mm by 6mm data cube is generated from a series of 200 B-scans, each composed of 200 A-scans. The disc edge is defined as the termination of Bruch’s membrane and the neuro-retinal rim width is determined by direct measurement of the neural tissue in the nerve head. By such means, the disc and rim area measurements correspond closely to the anatomy in the same plane as the nerve head and this approach also overcomes the distortion of the image which occurs when viewing a tilted disc. Software within the instrument then identifies the centre of the optic nerve head and evaluates the RNFL thickness along a 256 A-scan 3.46 mm diameter circle around the nerve head. When misalignment of the data cube occurs, the RNFL circle analysis is automatically corrected by the instrument software. Similarly, the thickness of the macular ganglion cell layer/inner plexiform layer complex is derived from one of two corresponding macular 6mm by 6mm cube scans (128 B-scans each composed of 512 A-scans and 200 B-scans each composed of 200 A-scans, respectively).
[CaptionComponent="2744"]
Standard automated perimetry
The most sensitive measure of the visual field is standard automated perimetry, ie, threshold perimetry generally using a white Goldmann size III stimulus presented on a white background of 10cdm-2 luminance.
Threshold perimetry estimates the differential light threshold, the minimum luminance (brightness), ?L, of the stimulus necessary to evoke a response presented against a background of given constant luminance, L. The output at each stimulus location is expressed in terms of the reciprocal of the threshold, ?L/L, the differential light sensitivity. The unit of measurement is the decibel (dB) where 0dB represents the maximum luminance of the stimulus and 1 dB is equivalent to a 0.1 log unit reduction in the stimulus luminance.
In the early 1990s, a number of studies suggested that between 25 per cent and 50 per cent of the retinal ganglion cells in POAG could be damaged before the manifestation of a visual field defect and that ganglion cell axons with larger diameter were preferentially damaged in glaucoma.9 This finding led to the development of alternative techniques of perimetry which purported to ‘target’ sub populations of retinal ganglion cells which were either preferentially damaged by the glaucoma process or exhibited lower degrees of overlap between adjacent receptive fields and would, therefore, demonstrate functional deficits earlier in the disease process since only a small number of cells would need to be lost prior to the loss of adequate retinal receptive field coverage.10 Such commercially available techniques of perimetry included Frequency Doubling Technology perimetry, short-wavelength automated perimetry, luminance pedestal flicker perimetry and critical fusion flicker perimetry. However, there is no clear evidence base to suggest that any of these alternative techniques determines visual field loss in advance of that by standard automated perimetry. Indeed, it is now accepted that the initial findings relating ganglion cell numbers to perimetric output were a consequence of the inadequacy of the histological process,11 of an initial comparison with the results of kinetic perimetry, and of a subsequent comparison with the results of threshold perimetry considered in dBs.12
The dB scale is logarithmic and, by definition, identical dB increments or decrements at different levels of sensitivity represent different increments or decrements of luminance when considered on a linear scale. Thus, the true nature of the structure and function relationship was initially confounded by the comparison of retinal ganglion cell count in linear units and the visual field outcome in logarithmic units. Such a comparison exhibits a curvilinear relationship which suggests, inadvertently, that, in early disease, structural damage is greater than functional damage and that, in more severe disease, functional damage declines at a greater rate than structural damage. However, when the two measures are considered on the same measurement scales ie logarithmic against logarithmic plots or linear against linear plots, the relationship between differential light sensitivity and ganglion cell number, and also other measures such as rim area, is linear over the abnormal range, at least. Indeed, it is now recognised that there is an almost one to one relationship between the number of ganglion cell soma calculated from standard automated perimetry and the number of ganglion cell axons calculated from OCT13-14 (Figure 4).
[CaptionComponent="2745"]
The printout for the optic nerve head and the RNFL characteristics generally presents quantitative measures displayed on a background colour which is coded for the statistical probability of the likelihood of the measured value lying within the age- and/or disc size/area-corrected normal range. Green indicates normality and yellow and red indicate abnormality at the p<5 per cent and p<1 per cent probability level, respectively. In addition, the print-out also generally displays two-dimensional B-scans of the optic nerve head and of the RNFL. The print-out for OCT generally contains the recording for both eyes.
A useful mnemonic for the interpretation of the optical coherence tomography print-out is WANDER. This was originally developed for the interpretation of the visual field print-out but can be also be applied to that from OCT, ie, What was done; are the results Accurate/Adequate; are the results Normal; if no, what type of Defect is present; Evaluate the type of defect (ie is it compatible with the signs and symptoms); Revaluate the defect (ie has it worsened)?
Interpretation of the RNFL and optic nerve head analysis report
The components of the printout for the RNFL and optic nerve head analysis report for the Cirrus HD-OCT are illustrated in Figure 5 but are applicable to most types of spectral domain OCT systems.
[CaptionComponent="2746"]
What was done?
This question refers to such factors as the given number and orientation of the A-scans forming the given number of B-scans of the given image, eg, the 200 by 200 optic disc cube (1), as highlighted in Figure 5.
Are the results accurate/ adequate?
With the Cirrus HD-OCT, the adequacy of the image is quantified by the signal strength which is measured on an integer scale from 0 to 10. The minimum acceptable signal strength is 6; however, a value of at least 7 or 8 is desirable.
After segmentation, the delineation of the layers in each B-scan image should be interrogated for the presence of errors whereby the image data is such that the instrument software is unable to differentiate, correctly, the various retinal layers either across the complete scan or in a localised region. Such errors may arise from epiretinal membranes and posterior vitreous detachments, with posterior staphyloma, and with extensive peripapillary atrophy, but may also be caused by gross movements of the patient during the image acquisition.
The print-out should also be interpreted with caution in the presence of posterior subcapsular and cortical cataract, which can lead to a reduction in the image quality.
The lateral dimensions of the en-face image are dependent upon the axial length of the eye. As an OCT instrument uses an average axial length, the lateral dimensions of the image may be over- or under-estimated in excessively shorter and longer eyes, respectively.
Is the scan normal?
If a defect is present, what type of defect is it?
Is the type of defect compatible with the signs and symptoms of the patient.
Quantitative measures of the optic nerve head parameters
The quantitative measures of the rim and disc area, the vertical cup-to-disc ratio and the cup volume (2) for each eye and between-eyes should be evaluated in the context of the qualitative impression of the optic nerve head obtained from slit lamp binocular indirect ophthalmoscopy. In the case of the Cirrus HD-OCT, these quantitative values are corrected for age and nerve head area; as a consequence, the disc area is not associated with any given statistical probability value. If the disc area is either smaller than 1.3mm2 or greater than 2.5mm2, respectively, a probability value for the remaining optic nerve head parameters cannot be calculated due to insufficient numbers of individuals within the normative data base manifesting such values. The range of normality for the vertical cup-to-disc ratio is surprisingly wide: in the case of a 69-year-old, for example, the normal range lies between 0.21 and 0.64.
Neuro-retinal rim thickness profile
The above values should also be compared with the Neuro-retinal Rim Thickness graph (3).
The Neuro-retinal Rim Thickness profile schematically displays, simultaneously in each eye, the neuro-retinal rim thickness at any given angular location referenced to the statistical likelihood of the value lying within the age- and disc area-corrected normal range.
The B-scan display (4) provides useful additional qualitative information not only of the depth of the cup but also of the corresponding slope of the cup walls; however, such information is not coded for probability level.
Quantitative measures of the RNFL thickness
The remaining part of the print-out is heavily weighted towards the interpretation of the RNFL thickness. The Average RNFL thickness (5) is corrected for age, only, and should be evaluated in the context of the corresponding disc size. Small and large discs will exhibit thinner and thicker peripapillary RNFL thicknesses, respectively, than the average size disc since the 3.46 mm diameter circle along which the measurement is made will be further from or nearer to, respectively, the maximum RNFL thickness.
The en-face RNFL thickness map
The en-face RNFL thickness map (ie the ‘straight ahead view’ of the RNFL as ‘seen’ by the clinician through the pupil) (6) is illustrated by false colours whereby red and other warm colours represent thicker values and cold colours represent thinner values with blue representing the thinnest values. The thickness map is analogous to the grey scale of perimetry; however, the shading of both maps is not corrected for age- or eccentricity-related normal variations in either thickness or sensitivity, respectively. The en-face thickness map in the normal eye should comprise warm colours, which become increasingly warmer as the vertical poles of the optic nerve head are approached, and represents the convergence towards, and entry into, the nerve head of the majority of the retinal ganglion cell axons.
The en-face RNFL deviation map
The en-face RNFL deviation map (7) illustrates the statistical probability of the given deviation lying within the normal range and the probability level is colour coded in yellow (<5 per cent) or red (<1 per cent) against a grey background. The RNFL deviation map is analogous to the Total Deviation Probability map for perimetry (see Part 2) and evaluates overall/generalised changes in the thickness of the RNFL.
The RNFL thickness ‘TISNT’ graph
The ‘TISNT’ graph (8) schematically displays, simultaneously in each eye, the peripapillary retinal nerve fibre thickness, ‘unravelled’ from the 3.46 mm diameter circle, in a profile format continuously from the horizontal temporal meridian to the superior, nasal and inferior meridians and back to the horizontal temporal meridian. The values are also referenced to the statistical likelihood of the given value at the given eccentricity lying within the age-corrected normal range.
The RNFL quadrant and clock hour average thickness pie charts
The average peripapillary RNFL thickness for each of the four quadrants, defined by the oblique meridia, and for each of 12 equal sectors (9) are referenced to the statistical likelihood of the value lying within the age-corrected normal range.
Case number 1
The patient in Figure 5 has POAG in the left eye and is a glaucoma suspect in the right eye. He exhibits, in the left eye, abnormal vertical and average cup-to-disc ratios (p<0.01 and p<0.01, respectively), an abnormal neuroretinal rim area (p<0.01) and an abnormal cup volume (p<0.05) (2). In the right eye, the vertical cup-to-disc ratio, only, lies outside the normal range (p<0.05) (2). The neuro-retinal rim thickness profile indicates that the rim thickness of the left nerve head exhibits focal thinning which is outside the normal range particularly superiorly but also inferiorly (3). The rim thickness is borderline normal in the right eye. The corresponding B-scans (4) show the extent of, and the difference in, the enlargement of the cup between the meridians and between eyes. The magnitude of the cup and the variations in the slope of the cup wall are particularly evident for the left eye. As might be expected from the characteristics of the left optic nerve head, the average RNFL thickness is outside the normal range (5) and is statistically thinner in the left eye than in the right eye. The RNFL thickness map (6) exhibits an absence of ‘warm’ colours at the superior and inferior poles of the left optic nerve head indicating a thinning of the nerve fibre layer in these regions. The red colour in the RNFL deviation map (7) indicates that the thickness of the RNFL in these regions lies outside of the normal range (p<0.01). In the right eye, the RNFL deviation map suggests a thinning of the RNFL in the papillo-macular region.
Thus, the information for the left eye, so far, indicates a predominantly vertically increased cup, with a consequent reduction in the neuro-retinal rim thickness and RNFL thickness, particularly superiorly but also inferiorly.
As would be expected from the above, the RNFL thickness ‘TISNT’ graph (8) for the left eye exhibits superior and inferior nerve fibre layer thicknesses which are abnormally thin (p<0.01 and p<0.01, respectively). The RNFL quadrant and clock hour (9) average thickness pie charts confirm the thinning at the superior and inferior poles. The thinning covers three sectors superiorly, two of which are at a probability level of p<0.05, while that inferiorly covers two sectors both of which are at a probability level of p<0.01. The values for the right eye would generally appear to lie within the normal range. The RNFL thickness in the right eye at the 9 o’clock sector (ie that which corresponds to the papillomacular bundle) is abnormal at the 5 per cent level and is a further illustration of the information displayed in the RNFL deviation map (7). Such an outcome is suspicious and requires follow-up.
Caution should be exercised in the context of what is becoming known as ‘red disease’ which is, essentially, a false-positive outcome ie, the occurrence of an abnormality due to chance. Such an eventually is known statistically as a Type I error: assuming a probability level of 5 per cent, such an error will occur five times in every 100 statistical comparisons. Conversely, the opposite, ‘green disease’ can also occur whereby manifest disease is not detected by the statistical process. Accordingly, it essential that the clinician has a sound knowledge of the various statistical outputs and ‘integrates’ the information with other clinical signs and with the symptoms.
Interpretation of the ganglion cell analysis report
The print-out for the ganglion cell analysis report derived from the 512 by 128 macular cube is illustrated in Figure 6 for a patient with POAG. The components of the print-out are based upon the same concepts as the print-out for the optic nerve head and RNFL thickness analyses. The patient in Figure 6 exhibits a gross reduction in the ganglion cell/inner plexiform layer thickness of each eye, as evidenced by the thickness maps, the deviation maps and the sectorial maps which are more pronounced for the left eye.
[CaptionComponent="2747"]
Case number 2
An interesting case is presented in Figure 7. The patient was being followed as an ocular hypertensive. At the most recent visit the peripapillary RNFL thickness of both eyes was within the normal range. However, the patient was found to have developed a lamellae macular hole in the left eye together with an epiretinal membrane. The visual acuity in the left eye was 6/7.5. Interestingly the ganglion cell/ inner plexiform layer thickness was grossly abnormal in both eyes and was indicative of POAG.
[CaptionComponent="2748"]
[CaptionComponent="2749"]
[CaptionComponent="2750"]
In Part 2, the author will discuss the interpretation of standard automated perimetry and look at how combined OCT and perimetric probability analyses represents a significant stride forward in the health assessment of the eye.
Model answers
(The correct answer is in bold text)
1 How does the axial resolution of a spectral domain OCT compare with the average retinal nerve fibre layer thickness?
A The RNFL thickness is typically ten times the axial resolution
B The two measurements are similar
C The axial resolution is typically ten times the average RNFL thickness
D Either might be greater than the other
2 How can disc measurement be standardised and repeatable in the case of anomalous, such as tilted, discs?
A The incidence of the light is adjusted
B The disc edge is always taken as the end of Bruch’s membrane and the neuro-retinal rim width taken from direct measurement of neural tissue in the nerve head
C The operator may subjective delineate the visible disc margin to compensate for error
D The neuro-retinal rim is assessed in the peripapillary zones to ensure standardisation
3 Which of the following statements about progressive retinal nerve disease is true?
A Functional loss is detectable as soon as structural loss occurs
B There is a linear relationship between structural loss when considered in linear unit and functional loss measured by logarithmic scales
C When similar scales (linear or logarithmic) are compared, there is near one to one relationship between ganglion cell soma calculated from SAP and ganglion cell axons from OCT
D Structural damage always exceeds functional loss until late stages of disease
4 The majority of patients in the Cirrus OCT normative database have which of the following ranges of disc area?
A 0.21 to 0.64 mm2
B 0.64 to 1.33 mm2
C 1.33 to 2.5mm2
D 0.21 to 2.5mm2
5 Which of the following statements about the RNFL thickness is true?
A It increases with age
B It increases with smaller disc size
C It increases with shorter axial length
D It is thicker in larger discs sizes
6 Which of the following statements about the ‘TISNT’ graph is true?
A It has peaks for the nasal and temporal meridians
B It is based on a 3.46 mm diameter circle centred at and surrounding the disc
C Values are represented irrespective of age
D Loss of thickness in glaucoma is likely to impact on the ‘valleys’ before the ‘peaks’
Professor John Wild is Professor of Clinical Vision Sciences at Cardiff University and Honorary Research Fellow at the University Hospital of Wales
References
A full list of references will be published in Part 2