Statistical Aspects of the Normal Visual Field in Short

Statistical Aspects of the Normal Visual Field in ShortWavelength Automated Perimetry John M. Wild,l Robert P. Cubbidge, 1 Ian E. Pacey,1 and Rosemary Robinson2 determine the intraindividual and interindividual characteristics of normal sensitivity derived by short-wavelength automated perimetry (SWAP) as a function of threshold algorithm. To determine also the influence of ocular media absorption on the magnitude of the interindividual variation in normal sensitivity, and hence the confidence limits, derived by SWAP. PURPOSE. TO

METHODS. The

sample comprised 51 normal subjects, stratified for age by decade (mean age, 555 years; range, 24-83 years) and experienced in white-on-white (W-W) perimetry and SWAP. One randomly assigned eye of each subject was examined on three occasions with Program 30-2 of the 640 Humphrey Field Analyzer using the Full Threshold and FASTPAC strategies for SWAP and W-W perimetries. Ocular media absorption (OMA) was assessed by the difference in scotopic sensitivity to stimuli of 410 and 560 nm. The group mean examination time (P < 0.001) was greater for SWAP than for W-W perimetry for both the Full Threshold (15.0% longer) and FASTPAC strategies (16.8% longer). The gradient of the age-decline in Mean Sensitivity for SWAP was approximately 25% less steep when corrected for OMA than when unconnected. The interindividual normal variability, expressed as the coefficient of variation, for SWAP without correction for OMA was 2.7 times greater (range 2.0-39), and with correction 1.9 times greater (range 1.4-2.9), than that for W-W perimetry.

RESULTS.

CONCLUSIONS. The

increased interindividual normal variability of SWAP, exacerbated by the lack of correction for OMA, currently limits the utility of SWAP in that the reduction in sensitivity required to indicate abnormality was proportionately greater than for W-W perimetry. (Invest Ophthalmol Vis Sci. 1998;39:54-63)

C

onsiderable interest is being shown in the development of short-wavelength automated perimetry (SWAP). SWAP preferentially stimulates the short-wavelengthsensitive pathway as opposed to the luminance pathway, which is examined by conventional white-on-white (W-W) perimetry. Several groups have demonstrated that SWAP is able to detect glaucomatous visual field loss before conventional W-W perimetry.1"7 SWAP has also been shown to yield more extensive visual field loss than W-W perimetry in optic neuritis,8 and laboratory studies have found that short-wavelengthsensitive pathway deficits precede luminance pathway deficits in diabetic maculopathy, age-related maculopathy, central serous choroidopathy, and retinitis pigmentosa.9"14 SWAP uses a two-color increment threshold procedure: a blue stimulus preferentially stimulates the short-wavelengthsensitive pathway, and a bright yellow broadband background simultaneously suppresses rod activity and adapts the mediumwavelength-sensitive and long-wavelength-sensitive pathways. The optimal stimulus and background parameters, which attain

From the ' Department of Vision Sciences, Aston University, Birmingham, United Kingdom, and Birmingham and Midland Eye Centre, City Hospitals NHS Taist, Birmingham, United Kingdom. Supported, in part, by the International Glaucoma Association. Presented, in part, at the 1996 annual meeting of the Association for Research in Vision and Ophthalmology, Fort Lauderdale, Florida. Submitted for publication June 23, 1997; revised October 2, 1997; accepted October 9, 1997. Proprietary interest category: C5. Reprint requests: John M. Wild, Department of Vision Sciences, Aston University, Aston Triangle, Birmingham B4 7ET, United Kingdom.

54

maximum isolation of the short-wavelength-sensitive pathway, are equivocal.12'5"715 However, the technique has now become commercially available on the Humphrey Field Analyzer (Humphrey Instruments, San Leandro, CA); a 440-nm narrowband blue Goldmann size V stimulus is presented against a 100-cdm~2 broadband (500-700 nm) yellow background.16 The size V stimulus and the 200-msec stimulus duration maximize spatial and temporal summation, respectively, thereby favoring greater isolation of the short-wavelength-sensitive pathway. Approximately 1.5 log units of response is obtained with these stimulus parameters before the loss of short-wavelength-sensitive isolation and the subsequent involvement of the medium-wavelength-sensitive pathway. The maximum dynamic range of SWAP is 65 apostilbs (20.6 cdm~2) compared with 10,000 apostilbs of W-W perimetry (3183 cdm~2). The standard algorithm for the determination of threshold in automated perimetry uses a 4-2 dB staircase strategy with a double crossing of threshold. In the case of the Humphrey Field Analyzer, the Full Threshold algorithm for the central field in W-W perimetry initially determines the threshold twice at each of four primary seed locations situated at 9° eccentricity in each quadrant. The starting luminance at these primary locations is 25 dB. The final 2-dB crossing of threshold can occur in either an ascending or descending direction, and the threshold is deemed to be the last seen stimulus luminance. Threshold is then determined at the immediate adjacent locations. The initial starting luminance at these secondary locations is 2 dB brighter than the expected value derived from knowledge of the sensitivity at the primary locations and of the slope of the hill of vision. The FASTPAC algorithm employs a single threshold crossing with a constant 3 dB step size, and Investigative Ophthalmology & Visual Science, January 1998, Vol 39, No. 1 Copyright © Association for Research in Vision and Ophthalmology

Normal Sensitivity in SWAP

IOVS, January 1998, Vol 39, No. 1

55

the threshold is taken to be the last seen stimulus luminance. The procedure for obtaining the threshold at the four primary locations and for calculation of the expected value at the secondary locations is identical with that of the Full Threshold algorithm. However, the initial starting luminance for the secondary locations is 2 dB dimmer than the expected threshold, when the expected value is an odd number, and 1 dB brighter, when the expected value is an even number. For SWAP, the two algorithms operate in the same manner with the exception that the starting luminance at each primary location is 20 dB. The FASTPAC algorithm for W-W perimetry offers an approximate 43% reduction in the examination time compared with the Full Threshold 4 -2 dB algorithm but at the expense of an increase in the intratest variability (short-term fluctuation) of approximately 25%.l7"19 The differences in mean sensitivity, examination duration, short-term fluctuation, and interindividual variability (and hence differences in the respective confidence limits for normality) between the Full Threshold and FASTPAC algorithms for SWAP, alone, and between SWAP and W-W perimetry are unknown.

media absorption for the 440-nm narrowband filter used in the commercially available SWAP will be greater than for the broadband filter, and the lack of a correction is likely to assume more importance. The influence of the lack of correction for ocular media absorption as a further source of increased interindividual normal variability for the 440-nm narrowband stimulus, in addition to that already present with SWAP, is unknown. The interaction of the type of threshold algorithm with the potentially wider confidence limits for the 440-nm SWAP is also unknown. The aim of the study was twofold: First, to derive the characteristics of normal sensitivity for the 440-nm narrowband SWAP, particularly the extent of any differences between the Full Threshold and FASTPAC algorithms, and also that between SWAP and W-W perimetry; and second, to determine the influence of the lack of correction for ocular media absorption on the magnitude of the interindividual normal variation in sensitivity derived by the 440-nm narrowband SWAP.

Fundamental to the delineation of visual field loss is the establishment of confidence intervals for normality at each stimulus location. Such intervals, based either upon Gaussian or empiric distributions, are used to identify abnormalities in the height and in the shape of the visual field.20 The normal visual field determined by SWAP is attenuated by ocular media absorption21 and by macular pigment absorption,21 which, respectively, cause a reduction in the height of the normal hill of vision and a depression in the foveal peak. The psychophysical determinations of ocular media absorption and macular pigment absorption are time consuming, particularly for the measurement of ocular media absorption, which requires the patient to be dark adapted. These determinations necessitate sophisticated visual judgements, and, consequently, are not clinically viable. The objective measurements of ocular media and macular pigment absorption require dedicated, expensive instrumentation and are beyond the scope of clinical practice. Clearly, therefore, any statistical interpretation package for SWAP must take into account the attenuation in sensitivity arising from the ocular media and from the macular pigment. Macular pigment absorption, although exhibiting large interindividual variability in the normal eye, is independent of age, influences only the immediate paracentral region, and, in the context of the detection and management of glaucomatous visual field loss using Humphrey Field Analyzer Programs 24-2 or 30-2, can effectively be ignored. However, ocular media absorption affects the whole field, increases with increasing age, and exhibits large interindividual variability in the normal eye for a given age. Previous studies of SWAP have corrected for the effects of ocular media absorption.1"7 The interindividual normal variability at each stimulus location for SWAP, derived with a broadband blue stimulus transmitting wavelengths below 475 nm, has been shown to be greater than for W-W perimetry.22 As a consequence, for a given significance limit, the deviation of the measured value from the normal value was greater for the broadband blue stimulus SWAP than for W-W perimetry.22 The magnitude of the SWAP interindividual normal variability, and hence the confidence limits, remained the same when the sensitivity was not corrected for ocular media absorption22; such a fortuitous finding indicated that the measurement of ocular media absorption in the clinical situation was unnecessary when using the broadband blue stimulus. However, ocular

METHODS The sample consisted of 51 normal subjects (24 men and 27 women) experienced in both conventional and short-wavelength Full Threshold automated perimetry with the Humphrey Field Analyzer. The sample was deliberately chosen such that the age was stratified across all decades. The mean age of the sample subjects was 55-5 years (SD 195), and the range was 24 to 83 years. Inclusion criteria comprised a visual acuity of 6/9 or better in either eye, distance refractive error less than or equal to 5 diopters equivalent sphere and less than 2.5 diopters cylinder, lenticular changes not greater than NCIII (Nuclear Color III), NOIH (Nuclear Opalescence III), CI (Cortical I), or PI (Posterior Subcapsular I) by Lens Opacities Classification System (LOCS) III,23 an intraocular pressure of less than 22 mm Hg in either eye, normal optic nerve head appearance, open angles, no history of a congenital color vision defect, no systemic medication known to affect the visual field, no previous ocular surgery or trauma, no history of diabetes mellitus, and no family history of glaucoma or diabetes mellitus. The study adopted a two-period cross-over design. The subjects were randomly assigned to one of four subgroups with each subgroup having four visits within a maximum interval of 2 weeks. At each of the first three visits, one randomly assigned eye of each subject was examined with the Humphrey Field Analyzer 640 using software version 931- The Humphrey Field Analyzer incorporated the standard commercially available hardware necessary for SWAP. At the first visit, all subjects underwent conventional W-W perimetry and SWAP using the Full Threshold strategy and Program 30-2 with the foveal threshold option enabled. One half of the sample was examined with W-W perimetry followed by SWAP, and the other half was examined with SWAP followed by W-W perimetry. Conventional W-W perimetry was undertaken using the default stimulus parameters of the Humphrey Field Analyzer; namely, a white stimulus, size III, presented for a stimulus duration of 200 msec on a 10-cdm~2 white background. SWAP was undertaken with the appropriate default stimulus parameters. At the second visit, the particular order of perimetric examination was maintained for the two groups. However, both groups were each randomly divided into two further categories. One subgroup from each main group was exam-

56

Wild et aL

IOVS, January 1998, Vol 39, No. 1

1. Group Mean and One Standard Deviation of Mean Sensitivity, Short-Term Fluctuation, Number of Stimulus Presentations, Examination Time, and Benefit-Cost Ratio for White-on-White Perimetry and ShortWavelength Automated Perimetry as a Function of Threshold Strategy TABLE

SWAP

W-W Perimetry

Mean sensitivity (dB) Short-term fluctuation (dB) Number of stimulus presentations Examination time (minutes) Benefit-cost ratio (clB~2)

Standard Strategy

FASTPAC Strategy

Standard Strategy

FASTPAC Strategy

28.0 (2.2) 1.29 (0.35) 461.2 (47.6) 14.0 (1-4) 0.0017 (0.0010)

27.9 (2.3) 1.61 (0.55) 262.2 (14.4)

23.1 (5.6) 1.60 (0.69) 508.1 (42.9) 16.1 (18) 0.0011 (0.0008)

23.7 (5.2) 1.89 (0.63) 297.6 (26.1) 9.5 (1.1) 0.0013 (0.0009)

8.1

(0.8) 0.0025 (0.0022)

Note that the dB scales between W-W perimetry and SWAP are not equivalent. * Abbreviations used: W-W, white-on-white; SWAP, short-wavelength automated perimetry; dB, decibels.

ined with the Full Threshold strategy and the other subgroup with the FASTPAC strategy. At the third visit, each subgroup was examined with the other strategy. The examination of the two visual fields at any given visit was separated by a 5-minute rest period. For SWAP, subjects looked into the bowl for 3 minutes before testing to ensure retinal adaptation of the medium- and long-wavelength-sensitive pathways. At the fourth visit, the ocular media absorption was determined in the designated eye at 15° eccentricity in each quadrant by measuring the difference in scotopic sensitivity to two narrowband (410 tun and 560 nm) stimuli. This method has become the standard for use with the Humphrey Field Analyzer.515'24"26 The background illumination of a modified Humphrey Field Analyzer 640 was extinguished, and the stimulus duration was altered to 100 msec using the perimeter software commands. Blocking material was fixed within the perimeter to reduce the leakage of light from the stimulus bulb onto the cupola. A red filter (Cinelux 406 red; Strand Lighting, Middlesex, UK) was positioned over the fixation target to preserve dark adaptation. Each subject was dark adapted for a minimum of 30 minutes. Fixation was monitored with the aid of an infrared source, which consisted of an M92, 20-watt, 12-volt, tungsten, halogen lamp mounted behind a Schott RG830 black light filter (transmitting wavelengths above 750 nm) (Osram, Middlesex, UK). The infrared assembly was encased within a steel tube and fixed to the frame of the perimeter such that the infrared radiation was reflected off the bowl surface and onto the subject. Scotopic thresholds for each wavelength were obtained three times at each of the four stimulus locations and expressed as a mean of the 12 measurements. The difference in mean scotopic sensitivity to the 4l0-nm and 560-nm stimuli was then scaled, using the equation of Norren and Vos,27 relative to the maximum intensity of the SWAP stimulus. The distance refractive correction, together with the appropriate near addition for the viewing distance of the perimeter cupola, was used in trial-lens form for each patient. The research followed the tenets of the Declaration of Helsinki and was approved by the Aston University Human Science Ethical Committee. Before undergoing examination, informed consent was obtained from each subject after the nature and possible consequences of the study, and the possible complications of the tests, had been explained. The visual field data corresponding to the Full Threshold and the FASTPAC strategies for conventional W-W perimetry

and for SWAP were expressed in terms of the visual field indices Mean Sensitivity and unweighted Short-term Fluctuation and in terms of the number of stimulus presentations, the examination time, and the staircase efficiency (the benefitcost ratio; defined as the reciprocal of the product of the number of stimulus presentations and the square of the shortterm fluctuation).1728 The two stimulus locations immediately above and below the blind spot were omitted from the analysis. The data for each of the five visual field parameters were subjected to a separate analysis of variance appropriate for a two-period, cross-over trial. The Mean Sensitivity, Short-term Fluctuation, and benefit-cost ratio were analyzed separately for conventional W-W perimetry and for SWAP, because the two perimetric measurement scales are not directly comparable. The maximum stimulus luminance associated with 0 dB differs between W-W perimetry and SWAP; however, each dB represents a 0.1 log unit change in stimulus luminance. The type of strategy (Full Threshold or FASTPAC), the type of perimetry (W-W or SWAP), the sequence of the two types of perimetry, and the sequence of the two strategies were all considered as intrasubject factors. The subjects and the age of the subjects were considered as interindividual factors. RESULTS The group mean and one standard deviation for the Mean Sensitivity, Short-term Fluctuation, number of stimulus presentations, examination time, and benefit- cost ratio for W-W perimetry and for SWAP, as a function of thresholding strategy, are given in Table 1. For W-W perimetry, the Full Threshold and FASTPAC strategies yielded similar group mean Mean Sensitivities (P = 0.790). The Mean Sensitivity declined with increased age (P < 0.001), and the extent of this reduction in sensitivity was similar for each strategy (P = 0.790). The results were also independent of the seqvience of the strategy (P = 0.836). The group mean Short-term Fluctuation for the FASTPAC strategy' was 24.8% greater than that for the Full Threshold strategy (P = 0.001), and this effect was independent of age (P = 0.524) and the sequence of the strategy (P = 0.215). For SWAP, the Full Threshold and FASTPAC strategies yielded similar group mean Mean Sensitivities (P = 0.065). The

Normal Sensitivity in SWAP

IOVS, January 1998, Vol 39, No. 1 Mean Sensitivity again declined with increased age (P < 0.001), and the extent of the reduction in sensitivity was also similar for each strategy (P = 0.857). The results were also unaffected by the sequence of the strategy (P = 0.217). The group mean Short-term Fluctuation for the FASTPAC strategy was 18.1% greater than that for the Full Threshold strategy (P = 0.006), and this effect was independent of age (P = 0.929) and the sequence of the strategy (P = 0.696). The group mean number of stimulus presentations was greater for SWAP than for W-W perimetry (P < 0.001). The group mean for the SWAP Full Threshold strategy was 10.2% larger than that for the W-W Full Threshold strategy, and the group mean for the SWAP FASTPAC strategy was 13.5% larger than that for the W-W FASTPAC strategy. The group mean number of stimulus presentations was greater for the Full Threshold strategy than for the FASTPAC strategy (P < 0.001). The group mean for the SWAP FASTPAC strategy was 41.4% less than that of the SWAP Full Threshold strategy, and the corresponding reduction for W-W perimetry was 43.2%. The reduction in the number of stimulus presentations for the FASTPAC strategy was similar between SWAP and W-W perimetry (P = 0.202). Age (P = 0.263), the order of the type of perimetry (P = 0.831), and the order of the type of strategy (P = 0.432) did not influence the number of stimulus presentations. The group mean examination time was longer for SWAP than for W-W perimetry (P < 0.001). The group mean for the SWAP Full Threshold strategy was 150% longer compared with that of W-W perimetry, and the group mean for the SWAP FASTPAC strategy was 16.8% longer than that for the W-W FASTPAC strategy. The group mean examination time was greater for the Full Threshold strategy than for the FASTPAC strategy (P < 0.001). The group mean examination time for the FASTPAC strategy was 41.2% shorter compared with the Full Threshold strategy in SWAP and 42.1% shorter in W-W perimetry. However, in absolute terms, the difference in the examination time between the Full Threshold strategy and the FASTPAC strategy was greater for SWAP than for W-W perimetry (P = 0.021). The group mean examination time increased with increase in age (P = 0.001), but this age effect was independent of the type of perimetry (P = 0.631) or the type of strategy (P = 0.087). The order of the type of perimetry (P = 0.831) and the order of the type of strategy (P = 0.432) did not influence the examination time. The benefit- cost ratio was more optimal in W-W perimetry for the FASTPAC strategy than for the Full Threshold strategy (P = 0.014), and this effect was independent of age (P = 0.649). For SWAP, the benefit-cost ratio was similar between the two strategies (P = 0.138) and did not vary with age (P = 0.269). The group mean ocular media absorption was 6.8 dB (SD, 1.42 dB). Ocular media absorption as a function of age is plotted in Figure 1. The group mean Mean Sensitivity, derived by SWAP and corrected for ocular media absorption, was larger than that without correction by 6.6 dB for the Full Threshold strategy and by 6.9 dB for the FASTPAC strategy. The decline in Mean Sensitivity with increase in age for both W-W and SWAP was described moderately well by a linear function with the coefficients of determination ranging from 0.33 to 0.51 (Fig. 2). For W-W perimetry, the slopes were 0.70 dB per decade for the Full Threshold strategy and 0.72 dB per decade for the FASTPAC strategy. For SWAP without correction for ocular media absorption, the slopes were 1.96 dB per

57

10 2- 9

8

I7 5

§

6

I 4s