Other psychophysical tests

Introduction

Glaucoma diagnosis and monitoring require assessments of both structure and function of the optic nerve and ganglion cell layer. The traditional tests of these parameters for the past half century or more have been structurally through ophthalmoscopy or photography and functionally by white light perimetry – either kinetic and, more recently, threshold static. Over the last few decades, these tests have been improved via computerized automation and/or laser-assisted quantitative structural analysis.Yet, the testing of both structure and function leave something to be desired. The assessment of structure by ophthalmoscopy and by stereo photography is subjective and varies greatly even among experts.1 Newer modalities of evaluating the optic nerve and ganglion cell layer using lasers and computerized analysis have greatly improved the subjectivity of this type of testing (see Ch. 14).

The ‘gold standard’ of modern functional assessment is automated threshold white-on-white perimetry (see Chs 9 and 10).While this type of testing is a big improvement in sensitivity, specificity, and repeatability compared to kinetic perimetry, significant problems are still present. Problems with this modality include tediousness for the patient, a significant learning curve, great variability even in experienced subjects from test to test, difficulty with interpretation,

and difficulty detecting both early damage and early progression (Box 11-1).2,3

One of the biggest problems with standard automated white-on- white threshold perimetry (SAP) is that it often does not show an abnormality until between 25% and 50% of the ganglion cells have

Box 11-1  Problems with automated threshold perimetry

Tedious

Not all patients can perform reliably Difficult for very young and for elderly

Instructions may be difficult for those not speaking the same language Time-consuming

Device not portable

Requires darkened and secluded environment Needs supervision

Not suitable for mass screening Variability from test to test Significant learning curve

Not sensitive to early glaucoma Difficult to detect progression

Cataract or other media opacity may confound results

been lost.4 In trained glaucomatous monkeys, about 50% ganglion cell loss must occur before significant white-on-white perimetric loss is seen.5 Thus, sensitivity of SAP to early glaucomatous damage is fair to poor. For some time now, researchers have been attempting to identify visual functions that may be more sensitive to early damage, either because the tests would be inherently more sensitive or because they target a subset of ganglion cells that by being fewer in number may have their specific function disturbed earlier than the more general functions of the total mass of ganglion cells.

There are several types of ganglion cells. Initially, it was thought that certain ganglion cells are damaged earlier in the glaucomatous process than other types.4 However, more recent studies suggest that all ganglion cells are damaged or killed by glaucoma but that some types of ganglion cells have less redundancy (fewer partner cells to cover for them if they become injured).6,7 Therefore, if a test can be targeted at ganglion cell functions that have little redundancy, the hope is that these functions will be affected early in the process. Short-wavelength automated perimetry (SWAP) and fre- quency-doubling technology (FDT) are two such tests (Fig. 11-1).

Another strategy is to find tests that have less noise (more reliable first tests with less variability between tests) than white-on-white (monochromatic) automated threshold perimetry. High-pass resolution perimetry and FDT are examples of this approach. Finally, attempts are being made using modern computer algorithms to either reduce the noise and variability or to better interpret from the standard test what is real and what is noise.The Swedish Interactive Test Algorithm (SITA) is one example of the former; neural network interpretation and the Glaucoma Progression Analysis (GPA) (Zeiss-Meditec, Dublin, CA) are examples of the latter. This chapter will discuss the most developed of these technologies, including SWAP, FDT, high-pass resolution perimetry, multifocal visual-evoked potential and multifocal electroretinogram. For a discussion of SITA and GPA, see the preceding chapters.

Color vision and short-wavelength automated perimetry

As understanding of the differential functions of the several types of ganglion cells became known, tests began to be developed that were targeted at specific ganglion cell functions that might have less redundancy than white-on-white perimetry. That color vision is affected in the glaucomatous process has been known for many decades.8 Acquired optic nerve disease affects the blue-yellow end of the spectrum as well as the red-green, whereas inherited defects tend to be limited to the red-green. So tests that include the

blue yellow end of the spectrum will be more specific for glaucomatous damage.

While cones are scattered throughout the retina, most are concentrated in the macula and surrounding area. Clinical color vision tests, either by design or accident, usually involve only the central cone-concentrated portion of the retina. The most commonly used color vision test is the Ishihara which does not effectively test the blue-yellow spectrum. Color vision tests that have been associated with early glaucomatous damage do evaluate, at least to some degree, that end of the spectrum and include the anomaloscope, Hardy-Ritter-Rand color plates, Farnsworth D-100, and Farnsworth D-15.8 Several studies have shown that color vision defects as demonstrated by these tests appear early in glaucoma, and may even precede defects seen on white-on-white perimetry which has been considered the gold standard for glaucoma diagnosis.9,10 Unfortunately for the clinical usefulness of color vision tests,

these instruments are insensitive, non-specific, too time-consuming or too difficult to interpret.8,11

It has been known for more than 80 years that the defects associated with acquired optic nerve disease such as glaucoma could be amplified on the tangent screen by using colored test objects as compared to the standard white test objects. However, tangent screen color testing was even more variable and dependent on ambient illumination, technical skill, and cleanliness of the test objects than white-on-black tangent screen testing. So, even though color vision is largely a central retinal function, enough color perception remains in the periphery to allow color perimetry. The Goldmann perimeter allowed for testing using colored test objects, but because of the technical skills required was rarely utilized.

The advent of computerized visual field testing gave us the possibility of eliminating or markedly reducing many of the variables that plagued manual perimetry. Early studies with SWAP determined that abnormalities detected by SWAP were predictive of ultimate white-on-white perimetry defects and that SWAP actually detected glaucomatous progression earlier than SAP.12 Abnormal SWAP results were found in ocular hypertensive eyes without white-on- white perimetry abnormalities that were at high risk for developing glaucoma, suggesting that the sensitivity of SWAP to early glaucoma

defects was higher than white-on-white perimetry. Abnormalities

found on SWAP in early glaucoma mapped well to abnormalities found in the structure of the optic nerve or nerve fiber layer.13–15

Eyes with normal SAP but with SWAP abnormalities were likely to have thin corneas and glaucomatous optic nerve appearance.16 Clearly, a characteristic glaucomatous abnormality on a SWAP test is likely to be correlated with glaucomatous optic neuropathy, abnormalities on laser scanning of the optic nerve and nerve fiber layer, and with thin corneas and, therefore, indicative of early glaucoma.

Early concerns that yellowing of the crystalline lens nucleus with age would reduce the function and require an age or lens color correction factor did not materialize.17 Fortunately, these observations made it possible to eliminate lens density testing and shorten the test to a clinically acceptable duration.

Short-wavelength automated perimetry tests the blue-cone mechanisms and the small bistratified ganglion (konio) cells that carry this information (SWS system). Several theories have been proposed for why SWAP should work. One is that the blue cones themselves, or their partner ganglion cells, are preferentially damaged in glaucoma; however, this is very unlikely since the patterns of ganglion cell loss are in the arcuate areas which do not correspond to the distribution of the blue cones or their ganglion cell partners. More likely is the fact that the K (konio-) ganglion cells that carry the information from the blue-cone system only represent a small proportion of the total ganglion cells ( 20%) and loss

of even a few of these cells would interfere with the total function since there is little functional overlap or cross-coverage.18,19

In the actual test performed on an optionally equipped Humphrey or similar automated perimeter, a blue-violet (440 nm), Goldmannsize V stimulus is projected onto a bright yellow (500 nm) background (Fig. 11-2). Note that the stimulus size is significantly larger than that usually used in white-on-white threshold perimetry (Goldmann size III). The brightness is gradually increased in randomly spaced trials until the stimulus is seen approximately 50% of the time.The determination of threshold in the standard SWAP test is the same as used in the full-threshold SAP. As of the writing of this chapter, the SITA, used so effectively in SAP, has been adapted to the SWAP test, allowing much shorter test times with less fatigue

Fig. 11-1  Ganglion cell types and functions. Some visual processing occurs in the retina (bipolar layer, etc.). The ganglion cells probably do not exclusively mediate the function assigned to them but they are largely responsible for carrying it to the brain. HRP, high resolution perimetry; SWAP, short-wavelength automated perimetry. (Courtesy of Chris Johnson.)

and less threshold variability.20–22 Since this algorithm is only a few months old as of this writing, experience is limited and little has been published, so most of this section, unless otherwise indicated, will be devoted to what is known about the standard, full-thresh- old SWAP.The output of the SWAP test is similar to the output of SAP (Fig. 11-3). STATPAC is available for SWAP with a norma-

Fig. 11-2  Short-wavelength automated perimetry test on a Humphrey automated perimeter. In the SWAP test, a blue (usually size V) test object is projected onto a yellow background, therefore isolating the blue-cone koniocellular ganglion cell system.

(Courtesy of Chris Johnson.)

tive database that can indicate the probability of abnormality at each tested point. Probably the most useful parameter for detecting early glaucoma with SWAP is the glaucoma hemifield test; it seems to be a sensitive and specific indicator of glaucoma-like damage.

Several independent, longitudinal studies have demonstrated the superiority of SWAP compared to SAP, both in early detection of abnormalities and in earlier detection of progression.23–28

Not all studies have shown that SWAP is more sensitive than

SAP.29 On the other hand, abnormal SWAP results correlate with optic nerve abnormalities even when the SAP is normal.30,31

Short-wavelength automated perimetry tests produce lots of noise, especially in naïve subjects, and a learning curve similar but actually longer than that seen for SAP has been demonstrated.32 Long-term fluctuation is greater with SWAP than with SAP.33 Defects seen on SWAP tend to be steeper and larger than those seen on either SAP or frequency-doubled perimetry.34

Patients with migraine also may show defects on SWAP testing;35 whether this represents a manifestation of independent nerve damage caused by the migrainous process or an association of migraine with glaucoma remains to be determined.Tamoxifen may induce abnormalities in SWAP (but not frequency-doubled perimetry) well before abnormalities are seen on SAP or in the retina.36

For interpretation, a SWAP can be considered abnormal if, on at least two exams, a pattern standard deviation is abnormal at worse

Table 11-1  Advantages and disadvantages of SWAP

than the 1% level, the glaucoma hemifield test is outside normal limits, there is one hemifield cluster with sensitivity below the 1% level, there are two hemifield clusters below the 5% level, there are four abnormal ( 0.05%) points, or there are five abnormal ( 0.05%) points on the pattern deviation plot.37 However, large variations are seen in ocular hypertensives examined with SWAP, and with different definitions of abnormality, large variations in who has abnormalities will be seen, suggesting that we do not yet have the right combination of diagnostic sensitivity and specificity to completely rely at least on a single SWAP test to determine if a true abnormality exists.38,39

In summary, SWAP is likely to detect functional glaucomatous damage and progression before SAP. However, problems with media opacities, length of test, fatigue, tediousness, high long-term fluctuation, and repeatability compared to SAP limit its usefulness, especially in the elderly (Table 11-1).40 It may be most useful in relatively young glaucoma suspects with good concentration for whom finding the earliest functional defect may play a role in determining management. As with any test, correlation of the results with other clinical findings is required to determine if the results of any single test or series of tests in a patient are reasonable and fit the clinical picture.41 The advent of SITA SWAP may reduce the tediousness, duration, and noise level of SWAP, but this remains to be independently verified.

Frequency-doubling perimetry

Frequency-doubling perimetry or technology (FDP, FDT) targets the function of a subset of ganglion cells – the M magnocellular ganglion cells that carry temporal information such as flicker and motion. These cells comprise only 10–15% of the total ganglion cells. There is little redundancy in this subset of cells and it would be expected that malfunction would be relatively easy to detect. In

fact, studies have shown that abnormalities in FDP, like those seen with SWAP, often precede those seen in SAP by several years.42,43

In this test, a low-frequency sinusoidal grating (0.25 cycles per degree) undergoes rapid reversal of light bars to dark 50 times per second (25 Hertz). The gratings are projected into different parts of the visual field and, at each location, projected at different levels of contrast between the light and dark bars (Fig. 11-4). The lower the contrast at which the grating is seen as a grating, the better is the sensitivity. Although the rapid reversal gives an optical illusion that the number of light and dark bars are twice as many in the same space as are actually present (doubled frequency – hence the name), it is not known if this optical illusion has anything to do

with the diagnostic ability of the test; in fact, the mechanism seems similar to that which detects any flickering stimulus.44,45 The origi-

nal device tested 17 sectors of the central 30°. Subsequent studies showed that even better sensitivity for glaucomatous defects was obtained with smaller gratings that were projected in a pattern similar to the 24–2 and 30–2 of the Humphrey automated perimeter. This latter observation formed the basis for the Humphrey Matrix (Carl Zeiss-Meditec, Dublin, CA) which is the newer, more versatile model.The 24–2 FDP testing strategy performs similarly to the 24–2 SAP in patients suspected of having glaucoma although the correlation is not one to one.46 Defects may be present on SAP that are not detected by FDP and vice versa.47

There are two general algorithms for testing similar to SAP.The screening mode is a threshold-related suprathreshold test. It is fast (1–2 minutes per eye) but not very useful for a monitoring baseline. The other algorithm is full threshold which takes closer to 10 minutes per eye. It is still faster than full-threshold SAP but no longer practical for screening. New algorithms such as the Zest program, which uses Bayesian logic, can cut the testing time in half without significantly affecting the accuracy.48,49 The ZEST program is available with the Matrix device and has made the threshold program with FDP quite practical and useful. The 20–1 threshold program of the FDP takes one-quarter to one-half the time of a SITA SAP without loss of specificity or sensitivity.50

In general, the FDP appears to be more sensitive and specific.46 Frequency-doubling perimetry has less test–retest variability than SAP, which might make it, theoretically at least, more sensitive at detecting changes over time.51 As noted above, defects on FDP may precede defects detected by SAP by several years. Furthermore, the defects on FDP in eyes with normal SAP relatively accurately predict the location of future SAP defects.52 Longitudinal studies do suggest that FDP is sensitive at detecting glaucomatous progression; however, the SAP and FDP do not always identify the same subset of patients, suggesting that patients may progress in different ways.53

Frequency-doubling perimetry has a faster learning curve than SAP (Table 11-2); furthermore, since the test is faster and the patients more comfortable with identifying the flickering target, fatigue is less of a factor.54 In the authors’ experience, patients overwhelmingly prefer the FDP to the SAP. While the learning curve is faster with FDP, the first attempt should not be relied upon in naïve subjects; however, the second attempt is usually accurate and stable.55 With the threshold programs, immediate retesting should be avoided as fatigue can be a factor after immediate repeat testing.56 However, in screening situations where the 2–minute screening program has been used, immediate repeat testing may help to significantly reduce false positives.

Frequency-doubling perimetry can be used reliably in children after 8 years of age57,58 but reasonable results may be obtained as

young as 5 years of age with proper training and preparation.59 From the beginning, it was shown that FDP was sensitive for

glaucoma defects.60 The results of FDP correlate well with SAP in both high-tension and low-tension glaucoma.61,62 Sensitivity and

specificity for all glaucoma defects are at about 90% compared to other visual tests for glaucoma.63 In particular, the sensitivity and

specificity for moderate to advanced glaucomatous defects are above 97%, which is an enviable record indeed.40,64 For early glaucoma,

the sensitivity is 85% and the specificity is 90%.64 Comparisons with SWAP and SAP are quite favorable for the FDP.65

Defects on FDP correlate with thin corneas in ocular hypertensive eyes, adding more evidence that both these parameters may

Table 11-2  Advantages and disadvantages of FDP

be risk factors for ultimate development of glaucoma as defined by SAP.66 Frequency-doubling perimetry also correlates well with structural abnormalities as defined by laser scanning devices such as the Heidelberg Retinal Tomography (HRT), Ocular Coherence Technology (OCT) and scanning polarimetry.67,68 However, the correlation is not perfect suggesting that the structure–function damage is not uniform from patient to patient.67

As happens with SAP, the screening programs which forego the careful evaluation of the thresholds are going to be less sensitive at detecting early glaucomatous visual field defects than the threshold strategies.69 With FDP, the 54–stimulus 24° field similar to the 20–2 of the Humphrey SAP perimeter (now included with the Matrix version) is somewhat more sensitive than the original 17–stimu- lus field.70 However, the 24–2 seems to be as sensitive as the 30–2 pattern so there is little gain in performing the longer test.69

Unlike SAP, FDP is relatively insensitive to refractive errors with up to about 6 diopters of spherical error having little effect on the