Short-Wavelength Automated Perimetry and Capillary Density in Early Diabetic Maculopathy

Abstract

purpose. To correlate short-wavelength cone-mediated sensitivity (SWS) assessed by blue-on-yellow perimetry with alterations of the perifoveal vascular bed in early diabetic maculopathy.

methods. Thirty-one patients (21 M, 10 F; mean age, 35 ± 12 years; no lens opacities) with no clinically significant macular edema were included in this study. All patients underwent short-wavelength automated perimetry (SWAP) and conventional white-on-white perimetry (Humphrey, 10-2). In digitized video fluorescein angiograms (Scanning Laser Ophthalmoscope), the size of the foveal avascular zone (FAZ) and the mean perifoveal intercapillary area (PIA) as a measure of capillary density were quantified interactively.

results. Mean thresholds of SWAP were significantly correlated with increasing size of FAZ (r = −0.51, P = 0.003) and PIA (r = −0.47, P = 0.01), whereas visual acuity expressed by log MAR (FAZ: r = 0.15, P = 0.41; PIA: r = 0.06, P = 0.76) and mean thresholds assessed with white-on-white perimetry (FAZ: r = −0.25, P = 0.20; PIA: r = −0.31, P = 0.14) were unrelated to diabetic changes of the perifoveal capillary network.

conclusions. The alterations of the perifoveal network are related to selective disturbances of visual function as measured by blue-on-yellow-perimetry. SWAP may act as an early detector of visual function loss in early diabetic maculopathy and serve as a helpful technique to predict early ischemic damage of the macula and to monitor therapy.

Visual field changes are commonly associated with diabetic retinopathy. ¹ Most studies have focused on proliferative and severe nonproliferative stages, when fundus alterations are clearly visible by ophthalmoscopy or fluorescein angiography. Thus, deep and large scotomas are associated with larger nonperfusion areas ^{2 3} and small scotomas with cotton-wool spots. ⁴ All these perimetric findings are confined to losses in the midperiphery in advanced disease. The loss of macular function, however, may be unrelated to the stage of retinopathy and remains the most common sight-threatening complication of diabetic retinopathy. New methods are required for understanding and defining diabetic maculopathy. Furthermore, new clinical tools are needed for screening in the early stages and managing further progression. 

The diagnostic problem of diabetic maculopathy consists in detecting very early morphologic and functional deficits related to later visual outcome. The established assessment of visual acuity does not have a high predictive value in the early stages, because acuity remains stable until approximately 55% of all neuroretinal channels are affected. ⁵ Morphologic changes assessed by fluorescein angiography, effective for detecting and quantifying capillary changes, are not reflected in visual acuity loss until the disease is well progressed. ^{6 7} Diabetic patients may exhibit an abnormally enlarged foveal avascular zone (FAZ) compared to healthy subjects, ⁸ without any measurable loss of visual acuity. Further psychophysical work-up, however, may reveal functional losses, particularly in color vision, ⁹ in patients with diabetic maculopathy and good visual acuity. A recent report on contrast sensitivity in diabetic maculopathy showed that neuroretinal function can be affected in the early stages and is related to alterations of the macular microvasculature. ¹⁰  

In this investigation, sensitivity was assessed in the central 10° visual field. In addition to the conventional white-on-white perimetry, short-wavelength automated perimetry (SWAP) was performed. The principle of this method is selective testing of the short-wavelength sensitive (SWS) cone-mediated mechanisms. This method is established in early detection of glaucoma, ^{11 12 13 14} where its use is to detect changes predominantly at the retinal ganglion cell level and loss of retinal nerve fibers. SWS mechanisms also are reported to be susceptible to damage in a variety of retinal diseases, ^{15 16} where changes are less specific for retinal nerve fibers and more confined to alterations of the inner retina. Animal experiments have shown a more selective loss of SWS cones to phototoxic or ischemic stimuli. ¹⁷ Several studies underlined SWAP abnormalities in diabetic patients with advanced retinopathy ¹⁸ or macular edema. ¹⁹ The purpose of the present study is to determine the association, if any, between retinal microcirculation and SWAP and conventional computerized perimetry and visual acuity in diabetic patients with normal visual acuity and without clinically significant macular edema. 

Materials and Methods

Patients

Prospectively, 31 patients with diabetes mellitus, visual acuity of 20/25 or better, and normal intraocular pressure (IOP) were recruited. Patients with a history of other eye diseases, particularly ocular hypertension and glaucoma, and eyes with a refractive error greater than 3 diopters (D) were excluded. In one patient with a physiologic large cupping, visual fields (Humphrey 24-2) and repeated IOP measurements were in the normal range. The clinical and demographic data are presented in Table 1 . The eye with the better visual acuity was selected; in case of equal acuity the eye with the lower refractive error was chosen. Findings of slit lamp examination of the anterior segment were normal in all eyes. Patients with nuclear opacities were excluded, in the interest of better quality of psychophysical and fluorescein studies. 

The fundus photographs of all the patients were classified as having no clinically significant macular edema according to Early Treatment Diabetic Retinopathy Study (ETDRS) guidelines. ²⁰ Retinopathy level was estimated by fundus photography according to ETDRS criteria. ²¹ Nine patients (29%) had no retinopathy (level 10), 5 (16%) microaneurysm formation only (level 20), 13 (42%) mild retinopathy (level 35), 1 had moderate (level 45), and 2 severe nonproliferative retinopathy (level 53), and 1 had proliferative (level 61) diabetic retinopathy. 

Best corrected visual acuity was determined by an ophthalmologist using objective refractometry (Rodenstock, Munich, Germany), lighting conditions, and standardized charts as described by DIN 58220, ²² followed by a complete ophthalmologic examination. For statistical analysis all visual acuity scores were converted into logarithmic equivalents (log MAR) for calculating acuity values or computing correlation coefficients. 

Informed consent was obtained from each subject, including detailed explanations of all procedures before participation in this study. The study protocol was reviewed and approved by the RWTH University Institutional Review Board for the use of human subjects. The tenets of the Declaration of Helsinki were followed. 

Methods

A Humphrey visual field analyzer (model 750; Humphrey-Zeiss, San Leandro, CA) was used for both blue-on-yellow (SWAP) and white-on-white (achromatic) conditions. The method was the standard procedure described in recent reports. ¹³ Stimulus size was Goldmann III (0.43° visual angle) for achromatic and Goldmann V (1.8° visual angle) for blue-on-yellow perimetry. The background was a 10 cd/m² broadband white for achromatic perimetry and a 200 cd/m² yellow. A full-threshold strategy was applied for the central 10° field (program 10-2). Limited StatPac (Humphrey-Zeiss) analysis (without probability data) was available only for the white-on-white condition. 

Quantification of the perifoveal capillary network was performed in fluorescein angiograms, as described in detail elsewhere. ^{8 23} The fluorescein angiograms were performed with a scanning laser ophthalmoscope (SLO; Rodenstock Institite, Ottobrunn, Germany; at 20° field). Within 20 to 50 seconds after dye injection, the capillary bed was well visualized, and sequences of 768 × 512 pixels images were digitized by a PC grabber card (Matrox Millennium; Matrox Graphics, Quebec, Canada). In one digitized image the perifoveal intercapillary area (PIA) and the foveal avascular zone (FAZ) were measured using digital imaging processing (Matrox Inspector, Matrox Graphics). PIA provides an estimate of capillary density in the network around the FAZ (5°). The FAZ and the perifoveal intercapillary areas were determined by an interactive image program. The borders of these areas were marked interactively by drawing around the surrounding capillaries with the cursor in the digital image. The area described by the cursor was calculated from the pixel size (5.6 × 5.6 μm) times the number of pixels within the boundary drawn by the cursor. PIA was calculated as the mean area of 100 measured single areas, randomly chosen in a 5° measuring circle around the center of the FAZ. Coefficients of variation of PIA[ cv(PIA)], calculated in one analysis characterize the homogeneity of the capillary vasculature. All angiogram analyses were performed in a masked fashion without prior knowledge of visual acuity, perimetry results, or determined EDTRS level. 

Reference values for comparison of the angiographic measures resulted from previous studies ⁸ and were derived from 31 healthy subjects of similar age (35 ± 9 years) examined in the same manner as the patients. 

Statistical Analysis

Mean value and SD are given for all samples with normal distributions (Kolmogorov–Smirnov test) and nonnormal distributions median and percentiles (2.5% and 97%). The unpaired Student’s t-test was used to assess the significance of the differences between groups. Findings smaller than 0.05 were considered statistically significant. Pearson correlation coefficients were calculated to evaluate the relationship between the parameters. P values were obtained after carrying out Fisher’s r to z transformations. 

Results

The study population consisted of patients with only mild macular changes. At least 65% had an FAZ size greater than the mean and 1 SD of the reference data. When compared to the age-matched normal subjects, the diabetic patients displayed significantly enlarged FAZ, whereas PIA did not differ significantly (Table 2) . The cv(PIA) was significantly higher in the diabetic population, indicating increased inhomogeneity of perifoveal vascularity. Figures 1 and 2 show examples of the perifoveal capillary network. 

Conventional white-on-white perimetry (achromatic) revealed a slightly reduced mean deviation and slightly increased pattern deviation (Table 3) . Because normative data are not available by the current software of the field analyzer, further statistics cannot be given. SWAP mean threshold (MT) had a higher range and differed more interindividually from white-on-white data (cvSWAP: 17%; achromatic perimetry: 6%). Achromatic perimetry MT (r = −0.30, P = 0.13) and SWAP MT (r = −0.27, P = 0.15) were not correlated to visual acuity expressed in log MAR. The SD was significantly higher (P < 0.0004) for SWAP than for achromatic perimetry. MT and SD for SWAP and achromatic perimetry were correlated to each other (MT: r = 0.64, P = 0.0002; SD: r = 0.42, P = 0.03). 

Figures 1 and 2 show the fluorescein angiograms of two patients, and Figures 3 and 4 contain prints of perimetric results for both achromatic and SWAP. MT and SD of SWAP were significantly correlated with the FAZ size and PIA, whereas visual acuity expressed by log MAR and mean thresholds assessed with achromatic perimetry were unrelated to changes of the perifoveal capillary network. Table 4 shows the correlation coefficients and P values, and Figure 5 shows MTs for both conditions plotted against the FAZ. 

SWAP MT may be influenced by transmission losses due to the aging lens. In this cohort there was no significant correlation of MT to the patient age (r = 0.07, P = 0.70). Furthermore, because SWAP SD is also correlated to capillary alterations, mean sensitivity losses do not seem to be strongly attributed to absorption of ocular media. 

Discussion

The angiographic measures performed in this study have proven to be of particular value in detecting early diabetic capillary nonperfusion even before microaneurysm formation occurs. ⁷ In advanced diabetic disease, reduction of capillary density is coupled with decreased visual acuity. ⁸ In this study, as reported previously, ¹⁰ enlarged FAZ was found even in patients with normal visual acuity and without clinically significant macular edema. Capillary alterations may result in tissue ischemia with subsequent macular edema, capillary dropout, development of microaneurysms, and neovascular complications. Changes of the perifoveal vasculature may be predictive of the visual outcome. 

A controversial issue in recent reports on diabetic maculopathy was whether a disturbance of neurosensory function or a change of the blood–retinal barrier is the earliest sign of pathology. ²⁴ Bek and Lund–Anderson ²⁵ reported that functional losses are relatively unrelated to localized blood–retinal barrier defects; however, macular capillary density, was not studied. 

Visual acuity is not sufficiently sensitive to provide clinical information about the impact of altered retinal function in early stages of diabetic eye disease. ²⁶ Frisén ⁵ found that visual acuity remains normal until approximately 55% of the neuroretinal channels are affected. Others have shown that contrast sensitivity is significantly reduced in patients with normal visual acuity in diabetic retinopathy. ^{10 26 27 28} Clinical studies have shown early color vision defects before diabetic retinopathy occurs. ^{29 30} Furthermore, a high spatial frequency loss of contrast sensitivity ^{10 26 27} points to an early involvement of parvocellular cells or cone-specific alterations in the diabetic disease course. 

Our study showed both loss of function and vascular change in early disease, when visual acuity is not affected. Alterations of SWS cone mechanisms, as found in this study, seem to be a more sensitive indicator of early visual impairment secondary to alterations in retinal vasculature. Further longitudinal studies must define critical thresholds for both methods and resolve whether functional losses precede morphologic alterations in diabetic retinopathy. Our results, particularly the increasing SD with increasing FAZ and PIA, are in agreement with recent studies in diabetic patients. Although Lutze and Bresnick ¹⁸ found no overall sensitivity loss compared to normals, significant sensitivity reduction and localized defects were detected with more severe diabetic retinopathy. Hudson et al. ¹⁹ found more localized visual field loss using SWAP compared to conventional perimetry in patients exhibiting a significant macular edema. 

The reason for the high susceptibility particularly of SWS cone mechanisms to ischemic processes remains speculative. ¹⁵ Early damage may be more readily detected in mechanisms with sparse neural representation as in SWS cone mechanisms. These have only a very small percentage of the total number of receptors and ganglion cells, ^{31 32} with receptive fields that do not overlap, ³³ compared to the longer wavelength cone mechanism. Also, differences in the response range ³² and the different nature of ganglion cells ³⁴ may account for earlier measurable function loss. Animal experiments indicate a higher vulnerability of SWS cones to phototoxic stimuli. ¹⁷ Johnson et al. ³⁵ found that SWS cone–mediated sensitivity (even after correction for ocular media) exhibit a loss with increasing age much more than the other cone systems, and they concluded there was a higher vulnerability of SWS cone mechanisms. 

The alterations of the perifoveal network are related to selective disturbances of visual function as measured by blue-on-yellow-perimetry. SWAP may act as an early detector of visual function loss in early diabetic maculopathy and as a helpful technique to predict early ischemic damage of the macula and to monitor therapy. For example, treatment of macular edema with focal photocoagulation is less effective if the procedure is performed after macular ischemia occurs. ³⁶ Particularly for understanding focal laser treatment and defining more accurate indications, functional alterations should be taken into account. 

In summary, the presence of perifoveal ischemia is related to a subtle deterioration of visual function as measured by SWAP. In patients without clinically significant macular edema and with normal visual acuity, SWAP could be a clinical adjunct for further identifying early ischemic diabetic maculopathy. 

 Supported by START 4/96–4 (RWTH Aachen).

 Submitted for publication May 14, 1999; revised August 9, 1999; accepted September 8, 1999.

 Commercial relationships policy: N.

 

Table 1.

 

Clinical and Demographic Data

No.	31 (21/10)*
Age (y)	35 ± 12
Intraocular pressure (mm Hg)	14.9 ± 2.9
Systolic blood pressure (mm Hg)	129 ± 17
Diastolic blood pressure (mm Hg)	78 ± 8
Hb A1c (%)	8.2 ± 2.5
Duration of diabetes (y)	11 ± 7

 Values are means ± SD.

*  Male/female sample size in parentheses.

Table 2.

 

Morphologic Parameters

Group	No.	PIA (μm2)	cv(PIA) (%)	FAZ (mm2)
Healthy subjects	31	3980 ± 587	35 ± 8	0.212 ± 0.078
Diabetic Study Group	31	4560 ± 1817	86 ± 20	0.270 ± 0.137
P value*	0.09	<0.0001	0.04

 Values are means ± SD. PIA, perifoveal intercapillary area; cv(PIA), coefficient of variation of PIA; FAZ, foveal avascular zone.

*  Unpaired Student’s t-test.

Figure 1.

 

Fluorescein angiogram (SLO 20° field, approximately 13° × 9°) of the perifoveal network (OD) of Patient A, a 28-year-old male diabetic patient (duration of diabetes: 20 years; mild nonproliferative diabetic retinopathy) with normal visual acuity (20/20). The image shows a relatively unaffected perifoveal capillary density with normal PIA (1900 μm²) and normal FAZ size (0.134 mm²).

Figure 2.

 

Fluorescein angiogram (SLO 20° field, approximately 13° × 9°) of the perifoveal network (OD) of Patient B, a 40-year-old male diabetic patient (duration of diabetes: 15 years; mild nonproliferative diabetic retinopathy) with best-corrected visual acuity of 20/20. Moderate loss of capillary density around the FAZ resulting in an increased PIA (4900 μm², above the mean and 1 SD of reference data) and FAZ (0.320 mm², above the mean and 1 SD of reference data) was detected.

Table 3.

 

Perimetric Results

	Mean (dB)	SD (dB)	Minimum (dB)	Maximum (dB)
WW MD	−1.51	1.84	−5.45	1.20
WW PSD	1.92	0.91	0.89	4.40
WW MT	31.06	1.80	28.07	34.18
WW SD	2.05	0.83	1.28	4.32
SWAP MT	24.81	4.21	12.68	29.49
SWAP SD	2.94	1.17	1.70	5.80

 WW, white-on-white perimetry; SWAP, short-wavelength automated perimetry; MD, mean deviation; PSD, pattern standard deviation; MT, mean threshold; SD, standard deviation.

Figure 3.

 

(A) Results of achromatic perimetry of Patient A (Figure 1) with no defects (MT: 33.1 dB, MD: +0.01 dB, PSD: 1.16 dB). (B) SWAP perimetry results from Patient A with no obvious defects (MT: 29.2 dB, SD: 1.70 dB).

Figure 4.

 

(A) Results of achromatic perimetry from Patient B (Figure 2) with no defects (MT: 31.2 dB, MD: −1.19 dB, PSD: 1.82 dB). (B) SWAP perimetry results of Patient B with subtle defects nasal to the center (MT: 24.1 dB, SD: 2.78 dB).

Table 4.

 

Correlation Coefficients of Perimetric Standard Parameters to Morphologic Data on the Perifoveal Vasculature

	FAZ (mm2)			PIA (μm2)			cv(PIA) (%)
		r	P	r	P	r	P
WW MD (dB)	−0.33	0.10	−0.36	0.09	−0.16	0.48
WW PSD (dB)	0.12	0.57	−0.02	0.94	−0.24	0.28
WW MT (dB)	−0.25	0.20	−0.31	0.14	−0.15	0.50
WW SD (dB)	0.11	0.57	−0.02	0.91	−0.27	0.21
log MAR (dB)	0.15	0.41	0.06	0.76	−0.19	0.33
SWAP MT (dB)	−0.51	0.003	−0.47	0.012	−0.34	0.07
SWAP SD (dB)	0.49	0.005	0.56	0.0015	−0.43	0.02

 WW, white-on-white perimetry; SWAP, SWAP: short-wavelength automated perimetry; MD, mean deviation; PSD, pattern standard deviation; MT, mean threshold; SD, standard deviation; PIA, perifoveal intercapillary area; cv(PIA), coefficient of variation of PIA; FAZ, foveal avascular zone.

Figure 5.

 

Mean threshold (MT) values as a function of foveal avascular zone: significant correlation of SWAP MT (□), no correlation of achromatic MT (•).