Quantitative Analysis of Retinal Vessel Attenuation in Eyes with Retinitis Pigmentosa

Abstract

Purpose.: Retinal vessel attenuation is a key finding in the diagnosis of retinitis pigmentosa (RP), but there have been no studies to determine whether quantitative measurement of this retinal sign is useful. We aimed to investigate retinal vessel caliber and its relationship with the severity of RP.

Methods.: This is a cross-sectional study based on 74 patients (145 eyes) with RP who had visual field assessment with Goldmann permeter and good-quality retinal images for vessel size measurements identified by retrospective medial chart review (1973–2007) in the electrophysiology clinic of a tertiary eye hospital in Australia. Retinal vessel calibers were measured using a computer-based program as the central retinal artery and vein equivalent (CRAE and CRVE). Goldmann visual field area for III4e white test light was measured quantitatively using ImageJ software as a clinical parameter to indicate the severity of RP. We used the generalized estimating equation models to estimate the difference in retina vessel calibers accounting for correlation between right and left eyes.

Results.: Mean CRAE and CRVE were significantly narrower in persons with smaller visual field area. For each 100 cm² decrease in visual field area, CRAE and CRVE decreased by −15.2 μm (95% confidence interval −20.7, −9.78) and −26.8 μm (−35.1, −18.5), respectively (both P < 0.001).

Conclusions.: In RP patients, the severity of visual field loss is correlated with retinal vessel attenuation. Quantitative retinal vessel caliber measurement may be a useful additional clinical marker for monitoring progression of RP or potential treatment response.

Introduction

Retinitis pigmentosa (RP) is a relatively common progressive retinal degeneration. ¹ RP is a leading cause of vision loss in young adults and affects 1 in 4000 people. ¹ RP consists of a heterogeneous retinal degeneration where patients typically develop poor night vision (nyctalopia), poor peripheral vision, and sometimes loss of central vision in later life. The progressive loss of rod and cone photoreceptors in the retina leads to gradual deterioration in visual function. RP may be inherited in autosomal dominant (30%–40%), autosomal recessive (50%–60%), X-linked (5%–15%), mitochondrial, and digenic forms with many causative genes identified. ¹  

Fundus examination shows the characteristic triad of bone spicule pigmentation, waxy optic disc pallor, and attenuation of retinal vessels. Retinal vessel attenuation is recognized as an almost universal finding in eyes with RP, ¹ and is mainly thought to reflect decreased metabolic demand of the degenerating retina, ² where loss of oxygen consumption leads to increased local oxygen levels in the inner retina that in turn results in vasoconstriction. ³ Retinal hemodynamic studies have found reduction in blood flow in RP eyes, ² whereas histopathological studies have reported sclerosis and atrophy of the retinal vasculature in extreme cases. ⁴  

Despite its importance as a diagnostic feature of RP, there have been few reports of quantitative retinal vessel measurement in eyes with RP ² and no correlation of this sign with the severity of the RP. Recent computerized techniques of measuring retinal vessel caliber have been extensively investigated as a marker of systemic vascular disorders, ⁵ such as diabetes, ^6,7 hypertension, ^8–10 and cardiovascular disease, ^11–16 as well as ocular conditions, such as glaucoma. ^17–19 We aimed to quantify the degree of vessel attenuation seen in patients with RP and determine whether the degree of visual field loss correlates with the degree of attenuation. The aim is to determine if quantitative retinal vessel caliber measurement may be a useful additional clinical marker for monitoring progression of RP or potential treatment response. 

Methods

Ethics Approval

This study had ethics approval from the Human Research and Ethics Committee at the Royal Victorian Eye and Ear Hospital in Melbourne and adhered to the tenets of the Declaration of Helsinki. 

Patients and Clinical Assessment

We identified 202 consecutive patients (354 eyes) with RP from archived fundus photography library between 1973 and 2007 in the electrophysiology and genetic clinic at the Royal Victorian Eye and Ear Hospital in Melbourne (Victoria, Australia). ²⁰ Of the 202 patients, 74 patients (145 eyes) had good-quality photographs for retinal vascular measurements and visual field assessment with III4e white test light with standard Goldmann permeter, and were included in this analysis. Before 1989, retinal photographs were performed with a Zeiss FF3 (Carl Zeiss Meditec, Jena, Germany), from 1989 to 1995 with a Kowa RC-WFV (Kowa, Nagoya, Japan) camera, and from 1995 onward with a TRC-50EX (Topcon, Tokyo, Japan) camera. Disc-centered photos were selected by an ophthalmologist, who scanned and measured retinal vessel calibers. Visual field area was measured from the result of III4e white test light of the Goldmann perimeter (Haag Streit, Bern, Switzerland). The visual field results were digitally scanned from the microfilms and the analysis was done using ImageJ software (National Institutes of Health, Bethesda, MD; available at http://imagej.nih.gov/ij/). ²¹ First, the magnification scale was calibrated across scans using the radius of the central 10-degree circle on the standard recording paper for the Goldmann perimeter, which is 1.2 cm as approximately 100 pixels. With this calibration, 100 pixels × 100 pixels area is equivalent to 1.44 cm² on the visual field recording paper. Then the scanned visual field result was displayed on the 12.1-inch (1400 × 1050 dots) display of the Thinkpad X61tablet laptop computer (Lenovo, Morrisville, NC). Using a digitizer pen, the visual field area for the III4e white test light was marked on the tablet screen; the marked area was measured using the “Measurement” function of the ImageJ software. 

Electroretinogram (ERG) response was determined based on the averaged scotopic b-wave ERG amplitude performed shortly after the field test. ERGs were performed following at least 25 minutes of dark adaptation using gold skin electrodes, Ganzfield Bowl with blue scotopic flash stimulation (Medilec, Melbourne, Australia). ERGs were categorized as “near normal” if within 95% of normative values, “subnormal” if outside 95% of normative values but response present, and “extinguished” if no response was measurable in scotopic ERG amplitude. Those with near normal ERG were patients with sectoral RP, very early in the course of the disease, and atypical RP. Scotopic light sensitivity levels were measured with the Goldmann-Weekers adaptometer (Haag Streit), and categorized into less than 10³, 10³ to less than 10⁴, 10⁴ to less than 10⁵, and 10⁵+ at 25 minutes. 

Retinal Vessel Caliber Measurement

Retinal vascular calibers were measured using a computer-based program (Retinal analysis – “IVAN,” University of Wisconsin, Madison, WI) following standardized protocols. ^22,23 In brief, a disc-centered fundus image was selected for each eye and used for the measurement. Before the measurement of retinal vascular caliber, we compared differences in magnification between camera types and calculated a magnification factor (i.e., image pixel to micrometer conversion) to calibrate the magnification of each camera type. This magnification factor was determined assuming an average one-disc diameter size is equivalent to 1850 μm. ²⁴ For each photograph, the largest six arterioles and venules coursing through an area one-half to one disc diameter from the optic disc were measured (Fig. 1) and summarized as the central retinal artery equivalent (CRAE) and the central retinal vein equivalent (CRVE) using formulas developed by Knudtson. ²² Intra-rater reproducibility was high with the intra-class correlation coefficient of greater than 0.85 assessed by using 100 standard images. Retinal vessel caliber measurements of selected cases are shown in Figures 2–4 with corresponding retinal image and visual field: Figure 2 shows a case with severe visual field narrowing and severe vessel attenuation in both eyes; Figure 3 shows a case with moderate visual field narrowing and moderate vessel attenuation in both eyes; Figure 4 shows a case with moderate visual field narrowing and vessel attenuation in the right eye, and severe visual field narrowing and vessel attenuation in the left eye. 

Figure 1. 

 

Identification and measurements of retinal arteriolar and venular caliber using semi-automated computer imaging program (University of Wisconsin, Madison, WI). Blue area in this image indicates selected retinal venular caliber.

Figure 2. 

 

Retinal images, visual field results (III4e) of a case with severe visual field narrowing and severe vessel attenuation in both eyes (A); ERG and scotopic sensitivity of a case with severe visual field narrowing and severe vessel attenuation in both eyes (B).

Figure 3. 

 

Retinal images, visual field results (III4e) of a case with moderate visual field narrowing and moderate vessel attenuation in both eyes (A); ERG and scotopic sensitivity of a case with moderate visual field narrowing and moderate vessel attenuation in both eyes (B).

Figure 4. 

 

Retinal images, visual field results (III4e) of a case with moderate visual field narrowing and vessel attenuation in the right eye, and severe visual field narrowing and vessel attenuation in the left eye (A); ERG and scotopic sensitivity of a case with moderate visual field narrowing and vessel attenuation in the right eye, and severe visual field narrowing and vessel attenuation in the left eye (B).

Statistical Analysis

All analyses were carried out using Stata (Version 12.1; StataCorp, College Station, TX). Differences in mean CRAE and CRVE were tested with the analysis of variance with Bonferroni-corrected post hoc analysis for category in age (5–20, 20–39, 40–49, and 60+ years old), age of onset (<20 or 20+ years), duration of the disease (<5, 5 to <15, and 15+ years), sex, and mode of inheritance (dominant, recessive, sporadic, and X-linked). We examined the association between visual field area, as a primary parameter used to indicate the clinical severity of RP, and retinal vessel caliber size in this study. Scotopic b-wave amplitude ERG response (near normal, subnormal, and extinguished) and scotopic sensitivity level at 25 minutes (<10³, 10³ to less than 10⁴, 10⁴ to less than 10⁵, and 10⁵+) were also used as secondary parameters. 

We used the generalized estimating equation (GEE) models to estimate adjusted mean difference in CRAE and CRVE accounting for correlation between right and left eyes. We used three models of (1) Model 1: simple nonadjusting model; (2) Model 2: adjusting for age, duration of the disease, and sex; and (3) Model 3: adjusting for age, duration of the disease, sex, mode of inheritance, and scotopic sensitivity. 

Results

Table 1 describes the overall clinical characteristics of study participants. Mean age was 35.0 (SD: 16.9) years old, and 56.8% were female. Median age of onset was 28.7 (SD 18.7) years old, and median duration of the disease defined by diagnosis date was 3.0 (interquartile range 1 to 10) years. Table 2 shows the associations between clinical characteristics and retinal vessel caliber. There was a decreasing trend in both arteriolar and venular calibers by duration of the disease (P for trend <0.001). Younger-onset patients (estimated age of onset <20 years old) had narrower retinal venular caliber than those with older age at disease onset (estimated age of onset 20+ years old) (P = 0.004). 

Table 1. 

 

Clinical Characteristics of Study Subjects (n = 74)

Characteristics			n	%
Sex	Female		42	56.8
Mode of inheritance	Autosomal recessive		33	44.6
	Autosomal dominant		15	20.3
	Sporadic		14	28.9
	X linked		8	10.8
	Not known		4	5.4
			Mean	SD
Age, y			35.0	16.9
Estimated age of onset, y			28.7	18.7
Estimated duration, y			3.0*	(1.0, 10.0)*
Visual Field area (III4e), cm2		Right	62.4*	(20.7, 118.2)*
		Left	67.4*	(21.5, 122.3)*
Scotopic sensitivity at 25 min	(unit)	Right	4 × 103*	(8 × 102, 8 × 104)*
		Left	8 × 103*	(8 × 102, 8 × 104)*
			n	%
ERG	Near normal	Right	5	6.8
		Left	5	6.8
	Subnormal	Right	41	55.4
		Left	42	56.8
	Flat or NA	Right	28	37.8
		Left	27	36.5
Vessel caliber measurement			Mean	SD
CRAE, μm		Right	101.9	20.7
		Left	101.0	20.3
CRVE, μm		Right	161.6	31.6
		Left	159.4	29.8

*  Median and interquartile range are shown.

Table 2. 

 

Retinal Vessel Calibers by Patients' Characteristics

Patient Characteristics		%	CRAE (μm)	SD	P Value	CRVE (μm)	SD	P Value
Age, y	5–<20	19.3	109.8	19.3	0.020	164.6	24.1	0.081
	20–39	42.8	97.6	17.7		155.4	29.1
	40–59	29.7	103.9	24.3		168.4	35.8
	60+	8.3	92.8	15.0		149.3	26.2
	P for trend		0.119			0.883
Estimated age of onset, y	<20	34.5	95.8	15.5	0.074	148.8	22.9	0.004
	20+	65.5	102.0	20.8		164.1	31.9
Estimated duration, y	<5	53.8	106.1	20.2	<0.001	169.9	29.8	<0.001
	5–<15	31.0	100.8	19.3		154.4	26.5
	15+	15.2	86.2	16.4		139.6	28.7
	P for trend		<0.001			<0.001
Sex	Female	57.2	103.0	18.8	0.295	160.3	26.4	0.938
	Male	42.8	99.4	22.5		160.7	35.6
Mode of inheritance	Autosomal recessive	44.1	97.9	21.1	0.005	157.9	34.4	0.467
	Autosomal dominant	20.7	99.9	15.5		156.0	25.4
	Sporadic	18.6	98.9	19.0		159.0	30.3
	X linked	11.0	113.5	20.7		166.3	28.1
	Not known	5.5	119.5	22.2		177.5	18.0

Both CRAE and CRVE were positively correlated with average visual field width (Spearman correlation coefficients = 0.40 and 0.47 for CRAE and CRVE, respectively [Fig. 5]). Mean CRAE and CRVE were significantly narrower in persons with smaller visual field area (P for trend <0.001) (Table 3). Persons with the narrowest quartile of visual field area had −15.2 μm narrower CRAE and −26.8 μm narrower CRVE compared with persons with the highest quartile of visual field area after adjusting in models 2 and 3 (both P < 0.0001). For each 100-cm² decrease in visual field area, CRAE and CRVE decreased by −15.2 μm and −26.8 μm, respectively (both P < 0.001) (Table 3). For each 1 SD decrease in CRAE and CRVE, visual field area narrowed by −29.3 (95% confidence interval [CI] −18.8, −39.8) cm² and −31.5 (−21.7, −41.3) cm², respectively (both P < 0.001). 

Figure 5. 

 

Retinal vessel caliber and visual field area (III4e) with fitted regression line with 95% confidence intervals. (A) Retinal arteriolar caliber and visual field area (Spearman correlation coefficient = 0.40). (B) Retinal venular caliber and visual field area (Spearman correlation coefficient = 0.47).

Table 3. 

 

Retinal Arteriolar and Venular Calibers and Visual Field Area

					Model 1		Model 2		Model 3
Visual Field Area (III4e)	CRAE (μm)	CRVE (μm)	SD	P Value*	Mean Difference	P Value	Mean Difference	P Value	Mean Difference	P Value
Per 100 cm2 decrease					−15.2 (−20.9, −9.49)	<0.001	−13.0 (−18.7, −7.29)	<0.001	−15.2 (−20.7, −9.78)	<0.001
Quartile 1 (0–20.7 cm2)	95.1		22.1	<0.001†	−16.2 (−24.8, −7.63)	<0.001	−13.1 (−21.7, −4.56)	0.003	−15.9 (−24.2, −7.64)	<0.001
Quartile 2 (21.4–66.7 cm2)	90.4		10.9		−20.9 (−29.6, −12.3)	<0.001	−19.3 (−27.8, −10.8)	<0.001	−19.7 (−27.5, −11.8)	<0.001
Quartile 3 (67.1–120.6 cm2)	108.2		18.8		−3.14 (−11.7, 5.38)	0.470	−4.31 (−12.6, 4.03)	0.311	−2.89 (−10.7, 4.88)	0.466
Quartile 4 (121.5–201.0 cm2)	111.3		20.7		(Reference)		(Reference)		(Reference)
Per 100 cm2 decrease					−26.7 (−34.9, −18.5)	<0.001	−23.9 (−32.2, −15.6)	<0.001	−26.8 (−35.1, −18.5)	<0.001
Quartile 1 (0–20.7 cm2)		144.1	28.4	<0.001‡	−31.0 (−43.6, −18.5)	<0.001	−27.6 (−40.2, −14.9)	<0.001	−31.9 (−44.7, −19.1)	<0.001
Quartile 2 (21.4–66.7 cm2)		148.0	21.2		−27.1 (−39.7, −14.5)	<0.001	−26.5 (−39.1, −13.9)	<0.001	−27.5 (−39.6, −15.3)	<0.001
Quartile 3 (67.1–120.6 cm2)		173.7	27.1		−1.45 (−13.9, 11.0)	0.820	−2.98 (−15.3, 9.35)	0.635	−2.66 (−14.7, 9.34)	0.664
Quartile 4 (121.5–201.0 cm2)		175.1	31.3		(Reference)		(Reference)		(Reference)

 Model 1: simple unadjusted model; Model 2: adjusting for age, duration of the disease and sex; Model 3: adjusting for age, duration of the disease, sex, mode of inheritance, and scotopic sensitivity.

*   P value for analysis of variance.

†  Bonferroni corrected post hoc analysis: Quartile 1 vs. 2: P = 0.99; Quartile 1 vs. 3: P < 0.001; Quartile 1 vs. 4: P = 0.002; Quartile 2 vs. 3: P = 0.001; Quartile 2 vs. 3: P = 0.020; Quartile 3 vs. 4: P = 0.99.

‡  Bonferroni corrected post hoc analysis: Quartile 1 vs. 2: P = 0.99; Quartile 1 vs. 3: P < 0.001; Quartile 1 vs. 4: P < 0.001; Quartile 2 vs. 3: P = 0.001; Quartile 2 vs. 4: P < 0.001; Quartile 3 vs. 4: P = 0.99.

There was a statistically significant decreasing trend in both CRAE and CRVE with altered ERG response or higher scotopic sensitivity threshold (P for trend for both CRAE and CRVE <0.001). Persons with extinguished ERG had significantly smaller CRAE than those with near normal ERG (−26.8 [95% CI −39.9, −13.8] μm and −34.7 [95% CI −55.5, −13.9] μm, respectively; both P < 0.001), and subnormal ERG (−10.3 [95% CI −16.9, −3.62] μm and −19.3 [95% CI −29.9, −8.60], respectively; both P < 0.001). Mean CRAE were significantly narrower in persons with subnormal ERG (−16.6 [95% CI −28.8, −4.37] μm, P = 0.008) compared with persons with near normal ERG. 

Mean CRAE and CRVE were significantly narrower in persons with higher threshold for scotopic sensitivity (P for trend for both CRAE and CRVE <0.001). Persons with the highest quartile of the scotopic sensitivity (≥8 × 10⁴) at 25 minutes had significantly smaller CRAE and CRVE compared with those with the lowest quartile of the scotopic sensitivity (<6 × 10²) threshold value (−13.0 [95% CI −23.3, −2.70] μm, [P = 0.013] and −19.5 [95% CI −35.8, −3.25] μm, [P = 0.019], respectively). Persons with the highest quartile of the scotopic sensitivity (≥8 × 10⁴) at 25 minutes had significantly smaller CRAE and CRVE compared with those with the second lowest quartile of the scotopic sensitivity (8 × 10² to 4 × 10³) (P = 0.018 for CRAE and P = 0.006 for CRVE). 

Discussion

Although retinal vessel attenuation has been recognized as one of the key features of RP, it has not been investigated quantitatively with regard to its association with the severity of the disease. The current study shows retinal vessel attenuation is associated with the severity of RP assessed as visual field width (Figs. 2–4). This association was independent of the effects of age, duration of RP, sex, and inheritance type. This was also supported by associations between altered ERG response or scotopic light sensitivity and retinal vessel attenuation. 

Retinal vessel attenuation is considered to be due to the decreased metabolic demand of the degenerating retina. ² In addition, loss of oxygen-consuming ozone receptors leads to increased oxygen levels in the inner retina, which in turn results in vasoconstriction. ³ Histopathological studies show a prominent layer of extracellular matrix (ECM) resembling the Bruch membrane deposited between the RPE and endothelin or cells of the thin walls of capillaries and venules. ⁴ In extreme cases, the ECM deposits completely occlude the vessel lumen with sclerosis and atrophy of the retinal vasculature that is seen clinically. It is thought that the thick perivascular deposits of hydrophobic lipid and elastin compromises the flow of nutrients from the vessel lumen to the neurons in the inner retina. ⁴ In addition, vasculature endothelial cells next to migrating RPE cells develop fenestrations that leak serum protein into the perivascular spaces seen clinically with leakage of fluorescein on angiography. Retinal hemodynamic studies show blood flow reduction by 76% in RP patients, although retinal vessels show regulatory responses to increased oxygen. ²  

Previous studies using quantitative measurements of retinal vessel caliber have shown that narrower retinal vessels are associated with a range of subclinical and clinical diseases not only in the eye but in other systemic organs. For example, narrower retinal vessels have been shown to be associated with subclinical cerebral infarctions ¹¹ and lacunar infarction, ¹² and predict incident clinical stroke, ^13,15 coronary heart disease in women, ^14,16 hypertension, ^8–10 and diabetes. ^6,7 Narrower retinal vessels are also associated with presence of glaucoma ^17,18 and reduced retinal nerve fiber layer thickness. ¹⁹ Our study therefore adds further data on the potential value of retinal vascular imaging for assessment of major clinical eye diseases. 

A strength of this study is using standardized retinal vessel caliber measurement, which enables us to quantify retinal vessel attenuation. This measurement takes on average only 5 minutes to complete for each retinal image, and it is highly reproducible. In addition, this study is based on a relatively large number of cases with RP at a tertiary hospital. Limitations of this study include its cross-sectional basis, and that we cannot determine whether retinal vessel attenuation precedes retinal degeneration or is due to retinal degeneration. Longitudinal follow-up study collecting concurrent retinal images and visual fields at multiple time points is warranted to elucidate the mechanism of retinal vessel attenuation in eyes with RP. Inheritance patterns were based on family history and examination of family members, thus it was not possible to comment on the detailed genetic subtypes of RP. Another limitation may be its basis in retrospective medical chart review, which we believe was not influential to our study results. Further prospective studies to investigate whether baseline retinal vessel calibers are associated with future progression of RP are warranted, with particular attention to RP genetic subtypes and more refined measures of electrophysiology. 

In summary, we describe a novel association between the severity of RP and retinal vessel attenuation assessed quantitatively from photographs using computer-imaging techniques. Thus, retinal vessel measurements may be potentially useful as additional objective and quantifiable clinical markers following the progression of RP and for monitoring potential treatments, such as vitamin supplementation ²⁵ or gene therapy, which has been used in Leber's congenital amaurosis. ^26,27  

Footnotes

 Supported by the Peggy and Leslie Cranbourne Foundation and the Pfizer Australia Senior Research Fellowship (DAM). The Centre for Eye Research Australia receives Operational Infrastructure Support from the Victorian government.

Footnotes

 Disclosure: Y. Ma, None; R. Kawasaki, None; L.P. Dobson, None; J.B. Ruddle, None; L.S. Kearns, None; T.Y. Wong, None; D.A. Mackey, None