Retinal vessel architecture and geometry are not impaired in normal-tension glaucoma

To investigate the associations between retinal vessel parameters and normal-tension glaucoma (NTG). We conducted a case–control study with a prospective cohort, allowing to record 23 cases of NTG. We matched NTG patient with one primary open-angle glaucoma (POAG) and one control per case by age, systemic hypertension, diabetes, and refraction. Central retinal artery equivalent (CRAE), central retinal venule equivalent (CRVE), Arteriole-To-Venule ratio (AVR), Fractal Dimension and tortuosity of the vascular network were measured using VAMPIRE software. Our sample consisted of 23 NTG, 23 POAG, and 23 control individuals, with a median age of 65 years (25–75th percentile, 56–74). No significant differences were observed in median values for CRAE (130.6 µm (25–75th percentile, 122.8; 137.0) for NTG, 128.4 µm (124.0; 132.9) for POAG, and 135.3 µm (123.3; 144.8) for controls, P = .23), CRVE (172.1 µm (160.0; 188.3), 172.8 µm (163.3; 181.6), and 175.9 µm (167.6; 188.4), P = .43), AVR (0.76, 0.75, 0.74, P = .71), tortuosity and fractal parameters across study groups. Vascular morphological parameters were not significantly associated with retinal nerve fiber layer thickness or mean deviation for the NTG and POAG groups. Our results suggest that vascular dysregulation in NTG does not modify the architecture and geometry of the retinal vessel network.

Normal-tension glaucoma (NTG) is a chronic neuropathy characterized by progressive change in the optic nerve head and subsequent visual field defects. The pathophysiology of NTG is complex and multifactorial. Intraocular pressure (IOP) is within statistical limits of normal (10-21 mmHg) however the level of IOP is a risk factor for progression 1 . Other potential risk factors have been identified including familial history, ethnicity, females, and vascular dysregulation [2][3][4] .
Investigators have reported an association between NTG, migraine and low systolic blood pressure, suggesting a possible role of vascular components in the pathogenesis of NTG 5 . Low blood pressure values measured during follow-up is correlated with structural degradation in medically treated NTG eyes 6 . Blood flow abnormalities have been reported in NTG, at the anterior part of the optic nerve head (ONH) 7 , the retrobulbar part of the optic nerve (ON) 8,9 , the choroid 10 and the retina 11,12 . Vascular risk factors have been also implicated in the pathogenesis and progression of POAG 13 , and are recognized to be predominant in NTG.
Fundus camera imaging is a non-invasive method for imaging the retina, often used for measuring parameters of the vascular retinal network. The vascular morphological phenotype (including tortuosity and fractal dimension) provides information on the architecture and geometry of the vessel network, which determines the efficiency of blood circulation. VAMPIRE (Vessel Assessment and Measurement Platform for Images of the Retina, Universities of Edinburgh and Dundee) is a suite of validated software tools enabling a quantitative analysis of the vascular morphometry 14 . VAMPIRE has been used in a plethora of previous studies, e.g., on lacunar stroke, cognition, dementia and cardiovascular diseases [15][16][17][18] , and recently to investigate the retinal vessel phenotype in ocular diseases, such as non-arteritic ischemic optic neuropathy and primary open-angle glaucoma (POAG) 19,20 .
The objective of this study was to compare retinal vascular parameters for patients with NTG versus POAG and controls after controlling for heterogeneity in the prevalence of comorbid conditions and glaucoma severity.
Sample size. From 82 eligible patients (during the period January 2020-April 2021), 23 NTG, 23 POAG and 23 controls subjects were included. We excluded a further 4 POAG, 2 NTG patients and 4 controls because of poor quality of the fundoscopic image or images were rejected by VAMPIRE. Three NTG patients were also excluded due to high myopia.

Data collection.
Medical history was collected on the basis of self-declaration, particularly cardiovascular risk factors, high blood pressure, diabetes, dyslipidemia and smoking. IOP-lowering medications were also reported. All participants underwent a complete ocular examination including visual acuity (VA, LogMAR), objective refraction (Tonoref 2™, NIDEK SA, Gamagori, Japan) and measurement of IOP (non-contact tonometry, TONOREF II, Nidek™, Gamagori, Japan). Anterior segment slit lamp examination, gonioscopy and fundus examination were performed to rule out other ocular diseases.
Study groups. Normal-tension glaucoma patients. The diagnosis of NTG used the following criteria: diurnal measurements of IOP ≤ 21 mmHg, characteristic optic nerve damage including cup-to-disc ratio > 0.5, rim thinning or notching; retinal nerve fiber layer (RNFL) and ganglion cell complex (GCC) defect; corresponding VF loss with abnormal results on the Glaucoma Hemifield Test and pattern standard deviation outside 95% of normal limits. All subjects underwent magnetic resonance imaging (MRI) of the brain and a carotid Doppler test to rule out secondary causes of optic nerve damage.
Primary open-angle glaucoma patients. POAG were matched 1:1 with NTG patients for age (within 5 years), refraction (for myopia: − 1 to − 3D; − 3 to − 6D with exclusion of high myopia > − 6D), systemic hypertension, diabetes, and mean deviation (MD) of the visual field (0 to − 6 dB; − 6 to − 12 dB; − 12 to − 18 dB). Inclusion criteria for the POAG subjects were a history of IOP > 21 mmHg, open-angle on gonioscopy, glaucomatous optic nerve damage, RNFL and GCC damage associated with corresponding visual field damages. Secondary glaucoma and eyes with an history of acute angle closure glaucoma were excluded.
Controls. Controls were matched 1:1:1 with NTG and POAG subjects for age (within 5 years), refraction (for myopia: − 1 to − 3D; − 3 to − 6D with exclusion of high myopia > − 6D), systemic hypertension and diabetes. Inclusion criteria for the control group were IOP ≤ 21 mmHg and normal ocular examination. Optical coherence tomography (OCT) examinations, including RNFL and GCC measurements, were normal in these patients.
Visual field and optical coherence tomography examinations. Patients with NTG and POAG underwent visual field and OCT examinations. This data could not be collected for controls as OCT is not part of the routine follow-up examination. Humphrey 24.2 sita-standard visual field parameters (mean deviation (MD), pattern standard deviation (PSD) and visual field index (VFI)) were recorded (Zeiss Meditec, Inc, Dublin, CA, USA). Visual fields with fixation losses lower than 20%, false-positive lower than 33% and false-negative lower than 33% were considered reliable and used for the analysis for all glaucoma patients.
Fundus image analysis. 45-degree funduscopic color photographs of both eyes were acquired, centered between the optic nerve and the macula after pupil dilatation, using two different non-mydriatic cameras: a Canon® CR-2 (30-2, Tokyo, Japan) with a resolution of 5184 × 3456 pixels and a Canon® CR-2 AF (9-1, Kawasaki, Kanagawa, Japan) with a resolution of 6000 × 4000 pixels. Funduscopic photographs were analyzed with the semi-automatic VAMPIRE software (version 3.2.0) by a single operator (ASL), after sitting the VAMPIRE operator training module.
First, the optic disc (OD) and the fovea were located, with manual correction when necessary. This enabled the definition of retinal coordinates (x axis through OD and macula centers, origin in the OD center) and circular www.nature.com/scientificreports/ concentric zones around the OD center, namely, zone A (between 0.5 and 1.0 optic disc diameter (ODD) from OD center), zone B (between 1.0 and 1.5 ODD from OD center) and zone C (between 1.0 and 2.5 ODD from OD center, Fig. 1). Vessels were subsequently detected and labeled as arterioles or venules automatically and corrected manually where necessary. CRAE and CRVE were computed as a weighted combinations of the widths of the six largest arteries and veins respectively using Knudtson's revised formulas 22 and arteriole-to venule ratio (AVR) as the ratio of CRAE to CRVE. The Fractal Dimension (FD) of the whole vascular network, a measure of the geometric complexity of the vessel network, was computed with the VAMPIRE implementation of a multifractal algorithm 23,24 . Arterial and venular vascular tortuosity was computed as the average of those of the six largest arteries and venules respectively) as defined elsewhere 25 . AVR, CRAE and CRVE were computed in zone B, fractal parameters and vascular tortuosity in zone C.
Raw measurements of CRAE and CRVE are computed in pixels by VAMPIRE. A pixel-to-microns conversion factor was obtained by dividing the average vertical ODD (over all images resized by VAMPIRE to a resolution of 3170 × 3170 pixels) by the assumed average of the disc diameter in microns (1850 μm), as previously described 22,26 . The conversion factor for the CANON CR-2 and CANON CR-2AF cameras was 3.9 microns per pixel.
Statistical analysis. Categorical data were reported as numbers and percentages and continuous variables summarized by median and 25th-75th percentiles. Comparisons across study groups were performed using the Kruskall-Wallis test for continuous variables and the Chi-square or Fisher exact test, as appropriate, for categorical variables. We used conditional logistic regression for paired samples to compare treatment categorical variables between NTG and POAG patients.
We used univariate linear regressions, modelling each retinal vessel parameter as a continuous dependent variable and RNFL and MD as continuous independent variables. We assessed the linearity assumption for continuous independent variables by using fractional polynomial functions. Two-sided p-values of < 0.05 were considered statistically significant. All analyses were performed using Stata Special Edition version 16.0 (Stata Corporation, College Station, TX, USA).
Twenty NTG patients (87%) were treated with at least one hypotensive topical medication (Table 2). Twentyone POAG patients (91%) were treated with at least one hypotensive topical medication ( Table 2). NTG cases were less likely to receive therapy with carbonic anhydrase inhibitors and had lower median PSD values than   3,4,5). No significant differences were observed in retinal vessel morphological parameters across study groups.
In NTG and POAG groups, no significant relation was found between CRAE, CRVE, AVR, tortuosity and fractal parameters with RNFL thickness or MD (Table 3).

Discussion
Using a matched nested case-control study design, no significant association was found between retinal vessel parameters estimated with VAMPIRE software and NTG.
When comparing NTG patients with controls, our results are consistent with a retrospective Indian study (n = 88 subjects including 28 NTG, 30 POAG and 30 controls) which reported no significant difference in retinal vessel diameters of affected quadrants by glaucoma according to the VF compared with that of unaffected quadrants in the same eye of age-and severity-matched NTG or POAG with hemifield damage 27 . In this latter study, vessels calibers were measured using Image J software at a distance of 1.72 mm of the center of the OD. In contrast, in previous studies using semi-automatic software conducted in Asian populations, NTG patients (n = 60) was associated with reduced arterial (109.8 ± 12 vs 120 ± 11.3 µm) and venular (158.5 ± 17.6 vs 176.8 ± 21.1 µm)     www.nature.com/scientificreports/ retinal vascular diameters compared to controls (n = 45, matched for age and sex) 28 . In the Singapore Malay Eye study (n = 3019 subjects including 74 NTG patients) 29 , logistic regression models adjusted for sex, age, smoking status, BMI, serum glucose, systolic blood pressure and hypertension status found that only narrower CRVE was associated with NTG. A Korean study 30 found that both CRAE and CRVE were reduced in unilateral NTGaffected eyes and unaffected NTG-fellow eyes as compared with controls. However, in this latter study, exclusion of patients or controls with systemic vascular diseases and systemic medications may represent a significant bias. We also found no significant association between CRAE and CRVE which VAMPIRE measures of arterial and venous retinal vessels with RNFL or VF indices in the two glaucoma groups, which is consistent with previous studies in NTG 27,30 or POAG 27 . This data is however in contrast with another study which found a significant correlation between RNFL thickness and CRAE in NTG patients (n = 60 previously untreated NTG) 28 .
A difficulty inherent to comparing results with previously reported ones is the difference in choices, e.g. retinal measurements to be used, algorithms to compute them, patient cohort characteristics and study design, among others. Within these limitations, a review of relevant work reveals the following. Firstly, the prevalence of NTG in Asians is known to be higher than in Caucasians 31 . Whether the severity of glaucoma, associated systemic and ocular vascular risk factors and genetic backgrounds are different in Caucasians and Asians is not known. Possible differences may explain discrepant associations between the retinal vessel phenotype and NTG. Secondly, some studies were conducted before glaucoma treatment 28,32 . The effect of hypotensive eye drops used by patients with glaucoma on retinal vessel diameter remains controversial. Dervenis et al. reported that the use of ophthalmic medications was associated with increased values of CRAE 33 . Conversely, Wong et al. found that antiglaucoma medications, in particular, topical beta-blockers, were associated with narrower arteriolar and venular diameters in individuals with and without arterial hypertension 34 . Thirdly, our study sample mainly consisted of patients with mild NTG (52.2% had MD < − 6 dB) and retinal vessel diameter may vary depending on glaucoma severity. As summarized in Table 4, visual field loss severity, non-available in some studies, varies with MD between − 4.7 and − 11 dB and RNFL thicknesses between 68 and 89.6 microns.
When comparing NTG with POAG patients, our study was consistent with two other studies which did not show any difference in the mean arteriolar and venular diameters between NTG and POAG 29,35 . Lee et al. reported smaller CRAE but comparable CRVE estimates in NTG relative to POAG eyes 32 .
In POAG, narrowing of arterial and venous diameter has been previously reported in Caucasians 20 and Asians 29 . Using the same Vampire software (version 3.0.3) 20 as in the present study, we reported that POAG was associated with a lower values of fractal dimension than controls but no difference for arterial and venous tortuosity. In the Singapore Malay Eye study, POAG was associated with decreased arterial and venous tortuosity and lower FD 36 . A prospective study showed that narrower CRAE was associated with a higher risk of occurrence of POAG, supported the concept that early vascular changes are involved in the pathogenesis of glaucoma 37 . The conflicting results between our previous study in POAG 20 and the present study may be due to different retinographs and lower resolution of images being used in the latter study whereas age and severity of glaucoma (MD, RNFL thickness) of POAG patients was similar. One should acknowledge that two patients with POAG benefitted from filtering surgery. However, the effect on IOP reduction after deep sclerectomy on CRAE is tiny (minus 6   Table 4. Primary studies reporting on retinal vessel parameters for normal tension glaucoma (NTG). Data are expressed as median (interquartile range (25-75th percentiles))* and mean (standard deviation)** AVR arteriole-to-venule ratio, CRAE central retinal artery equivalent, CRVE central retinal vein equivalent, ITA inferotemporal artery diameter, ITV inferotemporal vein diameter, MD mean deviation (dB: decibels), NA not available, RNFL retinal nerve fiber layer, STA superotemporal artery diameter, STV superotemporal vein diameter, TRT treatment.

Studies
Our study Chang et al. 28 Amerasinghe et al. 29 Shin et al. 30 Rao et al. 27 NTG population www.nature.com/scientificreports/ microns, median) 38 . POAG population significantly differed from the NTG population by a more frequent use of carbonic anhydrase inhibitors (CAI). Since CAI medication have not been reported to have a significant effect on retinal vessel diameters, 4/12/2023 11:58:00 AM this difference in using CAI drops between groups may not have influenced retinal vessel diameters. Finally, the vascular dysregulation associated with NTG is based on data from the retrobulbar vessels 8,9 , the anterior part of the ONH 7,41-43 , the retina 11,12,44,45 and the choroid 10 . These studies of different vascular ocular beds with different modes of regulations (myogenic, autonomic innervation) and different modes of explorations (Laser Doppler, Laser speckle flowgraphy, OCT-angiography, OCT-Doppler, fluorescein angiography, dynamic retinal vessel analysis, measurements of retinal vessels) suggest a vascular risk factor for NTG. Other risk factors were reported in the literature, such as low nocturnal blood pressure 46,47 , reduced ocular perfusion pressure [48][49][50][51] , vasospasm 52,53 , and migraine. 54,55 The absence of retinal vascular changes in Caucasian patients does not exclude vascular dysregulation since VAMPIRE measurements focus on the retinal vascular geometry only. These measurements do not enable one to explore vascular reactivity.
The strengths of our study include the high resolution of fundus images and study groups matched by age, refraction, systemic hypertension, diabetes and severity of glaucoma, all parameters which can have an impact on the retinal vasculature 56 . One trained grader (ASL) evaluated all the images (certified in the VAMPIRE training module).
Some limitations of this study should be discussed. First, our study might be underpowered due to the relatively limited sample size. Second, the observational design of the study does not allow to exclude unmeasured confounding factors. Third, medical history was self-reported and we did not use standardized criteria for cardiovascular risk factors, with the potential for misclassification. Fourth, our study was hospital-based: control and cases eyes were retrieved from the database of patients from the same department and a potential Berkson bias could not be excluded 57 .
In conclusion, we did not find significant differences between morphological parameters of the retinal microvasculature (arterial and venous vessels diameters, tortuosity, FD) between NTG, POAG and controls. No relation was found between retinal vascular parameters and MD or RNFL in NTG and POAG groups. These results underline that phenotype of the retinal vessel architecture and diameters using semi-automatic software from fundus images does not reveal an impact of the potential dysregulation of NTG.

Data availability
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.