Intervisit Repeatability of Retinal Blood Oximetry and Total Retinal Blood Flow Under Varying Systemic Blood Gas Oxygen Saturations

Abstract

Purpose: The aim of the study is to validate and calibrate the Doppler spectral-domain optical coherence tomography (SD-OCT) derived total retinal blood flow (TRBF) and metabolic hyperspectral retinal camera (MHRC) derived oxygen saturation (SO₂) of major retinal vessels in human volunteers using a novel and exact provocation technique (RespirAct) that allows the precise control of the end-tidal partial pressure of oxygen (P_ETO₂). Between visit repeatability of the TRBF and retinal blood SO₂ also were studied.

Methods: One eye of 11 young healthy subjects was chosen randomly for the study. Total retinal blood flow and retinal SO₂ measurements were obtained under conditions of normoxia, hyperoxia, and hypoxia. The order of hyperoxia and hypoxia was randomized between subjects. The measurements were repeated after a week interval.

Results: When the arterial P_ETO₂ was increased from baseline (P_ETO₂ = 100 mm Hg) to 200 and 300 mm Hg, the TRBF significantly reduced (P = 0.02) from 44.60 (±8.9) to 40.28 (±8.9) and 36.23 (±4.6) μL/min. Lowering the arterial P_ETO₂, from baseline to 80, 60, and 50 mm Hg, TRBF significantly increased (P = 0.04) from 43.17 (±12.7) to 45.19 (±5.5), 49.71 (±13.4), and 52.89 (±10.9) μL/min with simultaneous reduction in the arterial blood SO₂ content from 99.3% (±5.8) to 95.6% (±5.1), 89.6% (±2.8), and 83.3% (±3.9), respectively (P = 0.00). The coefficient of repeatability (COR) of TRBF, retinal arterial, and venous blood SO₂ values are 21.8 μL/min, 18.4%, and 15.2%, respectively.

Conclusions: Total retinal blood flow and retinal blood SO₂ measurements performed under safe levels of hypoxia and hyperoxia were repeatable in healthy adults over the range of SO₂ investigated.

The retinal blood vessels carry the necessary oxygen (O₂) and other essential nutrients to the retina, to meet its huge metabolic demand.¹ The delivery of O₂ to the retina is determined by factors, such as retinal blood flow (RBF) and the arterial/arteriolar blood O₂ content. The efficient regulation of blood flow and O₂ supply is vital to the retina to preserve vision. Therefore, we focused on measuring the amount of blood flowing through the retina and the O₂ dissolved in those blood vessels, to explain how the retina regulates overall blood and O₂ supply. 

The retina regulates blood flow in response to the local tissue demand. Despite the absence of an intrinsic sympathetic nerve supply, RBF can be maintained constant over a wide range of perfusion pressure.^2–6 This process is termed as “autoregulation.” The smooth muscle layer and the endothelial cells lining the blood vessel wall has a significant role in regulating the blood flow by enabling the constriction and dilation of the blood vessel, thereby decreasing and increasing the flow, respectively. The autoregulating ability of the retinal blood vessels has been reported previously.^7–15 Several studies have demonstrated changes in retinal vessel diameter, velocity, and flow to various provocative stimuli, such as carbon dioxide (CO₂), O₂ and flicker in healthy^12,16,17 and diseased cohorts.^18–21 

Hypoxia, a decrease in O₂ concentration, has been shown to vasodilate the retinal vessels.^22,23 Conversely, an increase in O₂ (hyperoxia) leads to vasoconstriction.^8,13 These changes occur to maintain constant O₂ delivery and to meet every day metabolic demands of the retina. Dysregulation of retinal vasculature is considered to be a precursor of major retinal diseases.^24,25 Impaired retinal vascular reactivity has been reported in diabetic retinopathy,^18,19 glaucoma,²¹ and also in smokers.^26,27 

The interest to develop imaging techniques to quantitate noninvasive RBF started in early 1970s,²⁸ when the first laser Doppler instrument was introduced to measure retinal red blood cell velocity. The application of Doppler technology in retinal imaging further evolved with the introduction of the Canon laser blood flowmeter (CLBF),^29–31 a technique that uses bidirectional laser Doppler velocimetry, and now as the functional extension of OCT, that is, Doppler SD-OCT^32,33 and bidirectional laser Doppler OCT.³⁴ Though the blood flow measurement, per se, may facilitate better understanding of major retinal diseases, including glaucoma, diabetic retinopathy, and age-related macular degeneration (AMD); more meaningful conclusion could be drawn if retinal blood oxygen saturation (SO₂) also could be measured. 

The introduction of spectral imaging techniques to detect and quantify molecular spectral absorbance profiles in retinal tissue is a major advancement in the field of retinal imaging. The Metabolic Hyper-spectral retinal imaging (MHRC) offers the potential to noninvasively quantify SO₂ disturbances in the retina. A plethora of evidence suggests retinal blood SO₂ changes in diseased eyes.^35–37 The novel application of Doppler SD-OCT and MHRC might offer the potential for early intervention, and insight into the pathogenesis of retinal pathologies.^33,38 

In this study, the SO₂ values of major retinal vessels and the total retinal blood flow (TRBF) are validated and calibrated in human volunteers using a novel and exact provocation technique (RespirAct) that allows the precise control of the end-tidal partial pressure of oxygen (P_ETO₂) to induce safe levels of hypoxia and hyperoxia. This technique uniquely targets exact P_ETO₂ and stabilizes the P_ETO₂ while maintaining isocapnia, irrespective of the individual participants' ventilatory response.^39,40 Also, the study interrelates TRBF and retinal blood SO₂ in healthy individuals. The P_ETO₂ was changed as defined by a series of step changes in inhaled oxygen and the reproducibility of SO₂ and TRBF values was assessed during normoxic conditions. 

Materials and Methods

Sample

This study was approved by the University of Waterloo Office of Research Ethics, Waterloo, and by the University Health Network Research Ethics Board, Toronto. One eye of 11 healthy subjects, mean age 33.36 years, SD 6.03 years was recruited. All subjects had a logMAR visual acuity of 0.0 or better. All participants were young, healthy, and nonsmokers. Exclusion criteria included any refractive error > ±6.00 diopters (D) sphere and/or ±1.50 D cylinder, IOP > 21 mm Hg, treatable respiratory disorders (e.g., asthma), systemic hypertension, cardiovascular disease, diabetes, endocrine disorders, medications with known effects on blood flow (e.g., antihypertensive medications with activity at autonomic receptors, smooth muscles, or those affects nitric oxide release), family history of glaucoma, or a history of any ocular disease. All the participants were asked to abstain from caffeine, red meat, and alcohol for 12 hours and avoid rigorous exercise approximately 1 hour before their study visit. Informed consent was obtained from each subject after a thorough explanation of the nature of the study and its possible consequences, according to the tenets of the Declaration of Helsinki. 

Study Visit

The study comprised two visits, with two sessions at each visit. During each visit, TRBF and SO₂ measurements were acquired under conditions of normoxia, hyperoxia, and hypoxia. The order of hyperoxia and hypoxia was altered across two visits separated by a week interval. Each visit lasted for approximately 3 hours. 

Instrumentation

Doppler Spectral-Domain Optical Coherence Tomography (SD-OCT).

The novel prototype Doppler SD-OCT uses the principle of “Doppler effect” to noninvasively quantitate the TRBF. The commercially available Optovue RTVue OCT (Optovue, Inc., Fremont, CA, USA), is a spectrometer-based OCT system, consist of a super luminescent diode with a center wavelength of 841 nm and a bandwidth of 49 nm. The axial resolution is 5.6 μm in tissue and transverse resolution is 20 μm. The scan protocol for TRBF measurement consists of double circular Doppler scans in the form of two concentric rings of diameters 3.40 and 3.75 mm centered on the optic nerve head, transecting all branch retinal arterioles and venules.³² A total of six scans were obtained and averaged for each ring. To measure the TRBF, the principle of Doppler effect is used where the frequency shift (Δf) from backscattered light is detected and simplified to  

 

where V is the velocity of the moving particle, θ is the angle between the OCT beam and the flow, n is the refractive index of the medium, and λ₀ is the wavelength of the incident beam. Flow in single vein = average (Vmax/Pi*Doppler phase shift*sin [Doppler angle])*vessel area, Vmax is the maximum speed corresponding to Doppler phase wrapping limit. Flow from individual retinal venules then is summed up to obtain the total volumetric blood flow at one time point. For more detailed explanation please refer to the study of Wang et al.³²  

From the measured Doppler shift within the vessel and Doppler angle estimation from the vessel center depth difference between two concentric rings, volumetric flow is derived using a semiautomated software (version 2.1.1.4) algorithm named Doppler optical coherence of retinal circulation (DOCTORC).^41,42 The repeatability of TRBF measurements acquired using Doppler SD-OCT was reported in previous publications from our laboratory.^43,44 

Metabolic Hyperspectral Retinal Camera (MHRC).

The Metabolic Hyperspectral Camera (Optina Diagnostics, Montreal, Canada) is a combination of a custom-built mydriatic fundus camera, a tunable light source, and a computer that controls image acquisition protocols, data storage, and data analysis. The fundus is sequentially illuminated using monochromatic light of predetermined range of wavelengths. At each wavelength, a 30° field-of-view of the posterior pole of the fundus is captured at high resolution (1.3 Megapixels). The filters are capable of delivering monochromatic light at a narrow bandwidth (Fourier transform microwave [FTMW] = 2 nm) and image acquisition occurs at a rate of 27 frames (wavelengths) per second. This allows the instrument to generate a stack of high resolution monochromatic fundus images (spectral data cube) within a few seconds. 

A spectral data cube obtained by the MHRC needs to be preprocessed before it can be analyzed. The spectral data cube is first normalized for spatial and spectral variations in light source intensity and any background “noise” generated from the system optics is removed. Next, each image of the data cube is registered spatially with the rest of the images in the stack to correct for any motion artifacts.⁴⁵ A preprocessed data cube then is opened with an in-house Matlab (The Mathworks, Natick, MA, USA) code. An automatic vessel segmentation algorithm⁴⁶ then is used to isolate the main vessel in the fundus image. The segmented vessel is analyzed further to determine the SO₂ (Fig. 1). In this study, SO₂ of a retinal blood vessel at half disc diameter distance from the disc margin was compared at different P_ETO₂ levels; images were captured between 500 and 650 nm in 5-nm steps for each stage of gas provocation. 

Figure 1

 

Oxygen saturation map of retinal vessels (scale 0%–100%). Some of the vessels show implausible changes in retinal blood SO₂ along their course. These artifacts are secondary to imperfections in image registration, but they have minimal effect on the calculation of blood SO₂ because the measurements are acquired within 1 DD of the optic nerve head where the vessels are relatively large and, therefore, impacted less by relatively small registration errors. The SO₂ measurement site was in this case on the inferior temporal arteriole and temporal to the optic nerve head.

Gas Delivery System.

A sequential rebreathing circuit (Hi-Ox⁸⁰; Viasys Healthcare, Yorba Linda, CA, USA) was used to provoke isocapnic hyperoxia and hypoxia. It comprises a fresh gas reservoir and an expiratory gas reservoir. Each reservoir is connected to a face mask with separate one-way valves. The face mask covers the mouth and nose of the subject. In turn, the two reservoirs are interconnected using a positive end-expiratory pressure (PEEP) valve, which allows subjects to breathe exhaled gas (i.e., rebreathe CO₂-enriched gas) when the fresh gas reservoir is depleted.^39,40 The subject's minute CO₂ production and O₂ consumption, gas flow, and composition entering the SGD breathing circuit was attained using an automated gas flow controller (RespirAct; Thornhill Research, Inc., Toronto, Canada) which is connected to a computer. The RespirAct has been described in detail in previous publications.^12,21 

Procedures

The study is performed over two visits. At the first visit, logMAR visual acuity and IOP (using the Goldmann Applanation Tonometer; Haag-Streit, Koniz, Switzerland) was recorded for both eyes. However, one eye was selected randomly for the study and dilated with one drop of tropicamide 1.0% ophthalmic solution (Alcon, Mississauga, Canada). Following that, a 10-minute resting time was given to the participants in a sitting position under room temperature to stabilize cardiovascular parameters. Participants were fitted with a face mask connected distally to the RespirAct face mask and sequential rebreathing circuit gas delivery system. At the end of this stabilization period, resting blood pressure, peripheral capillary oxygen saturation (S_pO₂), retinal blood SO₂, and TRBF measurements was taken during normoxia, hyperoxia, and hypoxia using the MHRC and the Doppler SD-OCT, respectively. 

The order of hyperoxia and hypoxia was randomized between subjects. Pulse rate, S_PO₂, and blood pressure were monitored continuously using a rapid response critical care gas analyzer (Cardiocap 5; Datex-Ohmeda, Helsinki, Finland) and transmitted electronically to a data acquisition system (S5 Collect, Datex-Ohmeda). A period of 10 to 12 minutes was given in between the gas provocation challenges. Visit 2 was conducted on the same participants, one week after the first visit, except that the order of provocation was altered this time. A diagrammatic representation of study protocol is illustrated in Figure 2. 

Figure 2

 

The figure demonstrates two different gas provocation protocols used for visits 1 and 2 under various P_ETO₂ levels (300–50 mm Hg). Protocol 1 (left): Isocapnic hypoxia (A–C) followed by isocapnic hyperoxia (D, E). Protocol 2 (right): Isocapnic hyperoxia (A, B) followed by isocapnic hypoxia (C–E). The order of provocation was randomized between subjects. BL, baseline.

Statistical Analysis

The normality of the data was ensured using Shapiro-Wilk test. The significant change in TRBF (μL/min) and arteriolar blood oxygen saturation (SaO2; %) and venular blood oxygen saturation (SvO2; %) during various gas provocation stages were analyzed using a repeated measures ANOVA (reANOVA). If a significant result was achieved using reANOVA, then post hoc testing was performed using Tukey's honestly significant difference (HSD) test. The coefficient of repeatability (COR) and coefficient of variability (COV) between visits was calculated as COR: 1.96*SD of difference; COV (%): SD/Mean*100. Bland & Altman plots illustrating repeatability of measurements between visits were plotted. For correlation analysis, Pearson's correlation coefficient (r) was used. An “r” can be any value between +1 and −1. A value greater than 0 indicates a positive association between the variables, whereas a value less than 0 indicates a negative association. Statistica software (StatSoft, Inc., Tulsa, OK, USA) version 12.0 was used for analyzing the data. The level of significance was set to be P < 0.05. 

Results

A total of 11 healthy subjects underwent retinal hemodynamic and retinal blood SO₂ measurements under conditions of normoxia, isocapnic hypoxia, and isocapnic hyperoxia. The participant's mean age was 33.36 years (±6.03). For the repeatability analysis, TRBF measurements from 11 subjects were included; however, for SaO₂ measurements, data of 10 subjects were included, except one, due to poor image quality. 

Systemic hemodynamic parameters for all participants across two visits are shown in Tables 1 and 2. Retinal hemodynamic parameters studied are given in Table 3. During both visits, the differences in heart rate (HR) and S_pO₂ at various P_ETO₂ levels, reached statistical significance (P < 0.05). Diastolic blood pressures (DBP) only showed a significant change (P = 0.007) during the first visit, while P_ETCO₂ and systolic blood pressures (SBP) were not different at the two study days. 

Table 1

 

Group Mean (± SD) for Gas and Cardiorespiratory Parameters Across Various Levels of P_ETO₂ During Visit 1

 

Table 2

 

Group mean (± SD) for Gas and Cardiorespiratory Parameters Across Various Levels of P_ETO₂ During Visit 2

 

Table 3

 

Group Mean (±SD) for Retinal Hemodynamic Parameters Across Various Levels of P_ETO₂ During Visits 1 and 2

 

Intervisit Repeatability and Variability of TRBF/Retinal Blood SO₂

The intervisit repeatability of TRBF and SO₂ measurements were analyzed and plotted using the Bland and Altman method as shown in Figure 3. The overall COR for TRBF, SaO₂, and SvO₂ measurements was 21.8 μL/min, 18.4%, and 15.2%, respectively. The overall COV for TRBF, SaO₂, and SvO₂ measurements was 15.1%, 4.7%, and 6.9%, respectively (Tables 4, 5). 

Figure 3

 

Bland and Altman plots showing difference in measurements as a function of average TRBF (left) and average SaO₂ (right) across the two visits. The dotted lines represent the limits of agreement and the center bar represents the mean of the differences between visits.

Table 4

 

Baseline Comparison of Mean, SD of Blood Flow, and Retinal Blood SO₂ Parameters Between Visits

 

Table 5

 

COR and COV for TRBF and Retinal Blood SO₂ for All P_ETO₂ Stages Between Visits

 

TRBF Response to Changes in P_ETO₂

Total retinal blood flow measurements during changes in P_ETO₂ are shown in Figure 4. There was no significant difference in baseline TRBF measurements as compared between visits. The average TRBF was 44.60 ± 8.9 μL/min during visit 1. When the arterial P_ETO₂ was increased from baseline (P_ETO₂ = 100 mm Hg) to 200 and 300 mm Hg, the TRBF significantly reduced (reANOVA, P = 0.02) from 44.60 (±8.9) to 40.28 (±8.9) and 36.23 (±4.6) μL/min, respectively. Conversely, lowering the arterial P_ETO₂, from baseline to 80, 60, and 50 mm Hg, increased the TRBF significantly (reANOVA, P = 0.04) from 43.17 (±12.7) to 45.19 (±5.5), 49.71 (±13.4), and 52.89 (±10.9) μL/min, respectively. A post hoc analysis for pairwise comparison was performed using Tukey's HSD test. The results show that the changes in TRBF was statistically significant only during baseline versus 300 mm Hg hyperoxia (P = 0.01) and baseline versus 50 mm Hg hypoxia (P = 0.04). 

Figure 4

 

Box plots represent change in TRBF at various P_ETO₂ levels. The legend in the middle of the box represents the median value, the upper and lower extremes of the box represent 25th and 75th percentiles. Error bars represent the nonoutlier range. Circles represent outliers. *P < 0.05; **P < 0.01.

SaO₂ and SvO₂ Response During P_ETO₂ Changes

Retinal blood SaO₂ and SvO₂ measurements during stable change in P_ETO₂ levels are shown in Figure 5. Lowering the arterial P_ETO₂ from baseline (100 mm Hg) to 80, 60, and 50 mm Hg, reduced the retinal arterial and venous blood SO₂ content from 99.3% (±5.8) and 56.3% (±4.2) to 95.6% (±5.1) and 52.5% (±4.1), 89.6% (±2.8), and 49.5% (±2.9), 83.3% (±3.9), and 45.0% (±6.1), respectively (reANOVA, P = 0.00). A Tukey's HSD test revealed a significant difference in SaO₂ and SvO₂ during baseline versus 80 mm Hg (P = 0.018, P = 0.013) baseline versus 60 mm Hg (P = 0.0001, P = 0.0003) and baseline versus 50 mm Hg (P = 0.0001, P = 0.0001). The SvO₂ was significantly different during baseline versus 200 mm Hg (P = 0.018) and baseline versus 300 mm Hg (P = 0.006). However, no significant change in retinal blood SaO₂ occurred at 200 and 300 mm Hg compared to baseline. 

Figure 5

 

Error bars show group mean retinal arteriolar blood SO₂ (left) and venular blood SO₂ (right) at various P_ETO₂ levels. *P < 0.05; **P < 0.01; ***P < 0.001.

Correlation Analysis of TRBF and SO₂

The relationship between dependent variables, such as TRBF, venous area, SaO₂, and SvO₂, during changes in P_ETO₂ was illustrated as scatterplots in Figure 6. The correlation analysis shows how the variables studied are associated with each other, for example, an increase in TRBF due to reduced P_ETO₂, decreases the retinal blood SaO₂ (Fig. 6A) and SvO₂ (Fig. 6B). The venous area changes, that is, vasoconstriction during hyperoxia and vasodilation during hypoxia, were correlated positively with a decrease and increase in TRBF, respectively (Fig. 6C). Using the Pearson correlation coefficient, a statistically significant relationship was found between TRBF and SaO₂ (r = −0.4, P < 0.05) as well as TRBF and SvO₂ (r = −0.37, P < 0.05) (Figs. 6A, 6B). Also, the correlation between retinal blood flow changes and simultaneous venous area changes during stable changes in P_ETO₂ was r = 0.5, P < 0.05 (Fig. 6C). 

Figure 6

 

Scatterplots of (A) TRBF (μL/min) against SaO₂ (%), (B) TRBF (μL/min) against SvO₂ (%), (C) TRBF (μL/min) against venous area (×10² mm²) during all gas provocation stages; dotted lines indicate confidence limits; different plot legend indicates various P_ETO₂ level. Squares, 300 mm Hg; hexagons, 200 mm Hg; diamonds, 100 mm Hg; triangles, 80 mm Hg; circles, 60 mm Hg; inverted triangles, 50 mm Hg.

Discussion

The current study showed that the RBF and retinal blood SO₂ measurements acquired using two novel prototype instruments, the Doppler SD-OCT and MHRC, during changes in arterial O₂ tension are repeatable and consistent. The previous methods to quantitate RBF, such as the CLBF and Laser Doppler Velocimetry (LDV), all are limited to quantitating blood flow from one vessel at a time.^34,47 Techniques, like fluorescein angiography, are invasive and have few associated side effects due to dye injection.^25,48 Ultrasound-based color Doppler imaging only determines the blood flow velocity; however, RBF quantification is not possible due to the lack of vessel diameter measurement.^48,49 

Overcoming the limitations of the aforementioned techniques, the Doppler SD-OCT could achieve TRBF from all major arterioles and venules. In this study, alongside the TRBF measurements, retinal arteriolar and venular blood SO₂ also was achieved using the MHRC. Recently, Palkovits et al.^22,50 reported retinal blood SO₂ and RBF during hypoxia and hyperoxia in humans. Retinal blood flow increased while retinal blood SO₂ decreased during two levels of hypoxia studied.²² During hyperoxia, a significant decrease in RBF, vessel diameter, and velocity observed as well as SO₂ increased in retinal arteries and veins by +4.4% and +19.6%, respectively.⁵⁰ One must note that, in their studies, RBF measurements were achieved from one single vein compared to TRBF reported in the current study. Also, the vessel diameter and blood flow measurements were acquired using two different instruments, unlike simultaneous acquisition using Doppler SD-OCT. 

The combination of computer-controlled gas sequencer along with the RBF and SO₂ measurements, allows the precise combinations of P_ETO₂, while clamping the P_ETCO₂ (i.e., to be able to achieve isocapnic hyperoxia and isocapnic hypoxia) concentrations, in turn this provides more reliable and reproducible data. 

Studies quantitating RBF and retinal blood SO₂ in humans are very few in the literature. Most of the experiments have equipped microspheres to provide direct measurements of inner retinal O₂ tension and O₂ consumption in animals. Such techniques are more invasive and less than ideal to be used in humans.^23,51 The current study validates two novel prototype instruments to measure RBF and SO₂ noninvasively in young healthy individuals. Validation of these techniques might facilitate further understanding of retinal O₂ extraction in normal as well as in diseased eyes. 

Our lab has reported previously a COR of 11% and 14% for SaO₂ and SvO₂, respectively, using a hyperspectral retinal camera in six healthy subjects.⁴⁵ The current study documents COR of 18.4% and 15% for SaO₂ and SvO₂, respectively, using MHRC. It is interesting to note that the COR for SvO₂ is similar to what the previous investigator has reported; however, the COR of SaO₂ has a greater difference compared to the previous study. 

It is unclear whether this could be due to larger variability in SaO₂ measurements among individuals. In our study itself, we noticed few subjects had SaO₂ values beyond 100% during baseline conditions (i.e., normoxia), which is beyond the physiological range reported in the literature.^52,53 Similar results also have been reported previously by many investigators. Mordant et al.⁵⁴ reported mean SaO₂ of 104.3% (±16.7%) in retinal arterioles using a “snapshot” hyperspectral spectral imaging technique. Hardarson et al.⁵⁵ achieved SaO₂ values ranging between 93% and 108% by using automated image analysis software to derive optical density ratios of arterioles. 

In contrast to flash or snapshot hyperspectral retinal cameras, the MHRC constructs a spectral data cube based on sequential imaging (nonflash). There is evidence to suggest that using flash illumination may artificially alter the measured retinal SO₂ values.⁵⁶ On the other hand, sequential imaging is more susceptible to motion artifact; however, MHRC's high frames per second imaging capability helps to minimize this effect. 

Hammer et al.⁵⁷ reported average SaO₂ and SvO₂ of 98% ± 10.1% and 65% ± 11.7%, respectively, under normoxia. During 100% O₂ breathing, the arterial and venous SO₂ increased by 2% and 7%, respectively. Hardarson et al.⁵⁵ have shown that the SaO₂ increased from 96% (±9%) to 101% (±8%) during hyperoxia. In our study, the SaO₂ and SvO₂ during normoxia was 99.3% (±5.8%) and 56.3% (±4.2%). During hyperoxia (P_ETO₂ = 300 mm Hg), the retinal SaO₂ and SvO₂ increased by 4.7% and 4.8%, respectively. In contrast, during hypoxia (P_ETO2 = 50 mm Hg) we found a reduction in SaO₂ and SvO₂ values to 16% and 11.3% compared to baseline. The variability of SO₂ measurements reported in our study is much less compared to those reported by similar studies in the literature. Figure 5 shows almost a “linear” trend of SaO₂ and SvO₂ in response to decrease in P_ETO₂ below 100 mm Hg. 

Garhofer et al.³⁴ reported a high interindividual variability in TRBF (44.0 ± 13.3 μL/min) measurements using bidirectional LDV in young healthy subjects. One must note here that the TRBF reported is not from the simultaneous measurement of retinal blood velocity and vessel diameter, rather it is derived from one vessel at a time due to the technologic limitations. In contrast to bidirectional LDV, Doppler SD-OCT uses the “Doppler shift” principle to quantitate the red blood cell velocity from all major arterioles and venules in a single point of time.³² Venous area measurements are extracted from the acquired Doppler OCT images using a semiautomated software named DOCTORC. Although manual steps are involved in venous area estimation using DOCTORC, few studies actually have reported the repeatability and variability of the manual grading technique, per se.^41–43 

A recent study from our lab has reported a COV of 7.5% and COR of 6.43 μL/min in young adults using Doppler SD-OCT.⁴⁴ Wang et al.^32,33 reported a mean TRBF in healthy young subjects of 45.6 ± 3.8 μL/min and COV of 10.5% using a single-beam Fourier domain (FD)-OCT. Our study reported a TRBF of 43.59 ± 9.2 μL/min, which is comparable to the previous studies. In our study, the reported COV and COR for TRBF during normoxia was 15.1% and 21.8 μL/min. During hyperoxia (P_ETO₂ = 300 mm Hg) TRBF was decreased by 8.37 μL/min and during hypoxia (P_ETO₂ = 50 mm Hg) TRBF increased by 9.72 μL/min compared to baseline. Due to the large variability in blood flow data (Fig. 4), as well as smaller sample recruited, a significant difference in TRBF was not achieved during the rest of the P_ETO₂ stages (i.e., P_ETO₂ of 200, 80, and 60 mm Hg). 

Hyperoxia is an increase in arterial partial pressure of oxygen from baseline homeostatic levels. Several studies have shown that retinal vessels react to hyperoxia by local constriction of arterioles, venules, and capillaries, thereby reducing the RBF.^8,13,15 Recently, Palkovits et al.⁵⁸ have reported the effect of breathing 100% oxygen on flicker-induced vasodilation in humans. In contrast to animal studies,⁵⁹ breathing oxygen was showed to increase the flicker-induced retinal vasodilation in humans. In our study, we mainly emphasized on the physiological responses of retinal vasculature to hyperoxia alone without involving neurovascular coupling. 

Studies from other labs have just used 100% O₂ or coadministered O₂ (∼>90%) and CO₂ (∼5%), without clamping the PCO₂ for hyperoxic provocation.^8,10,16 This might further reduce the PCO₂ concentration, which might impact the measured variables.⁶⁰ The novel computer-controlled gas provocation technique used in the current study has overcome the aforementioned limitation by minimizing alterations to the systemic PCO₂ concentration during hyperoxic and hypoxic provocation, thereby streamlines the retinal vascular reactivity response to O₂ only. It has been reported that endothelin-1 is known to mediate the vasoconstrictive response to hyperoxia.⁶¹ However, animal studies have reported that other factors, such as thromboxane and 20-hydroxyeicosatetraenic acid, might as well contribute to hyperoxia-induced vasoconstriction.⁶² This remains to be investigated in humans. 

Hypoxia leads to increase in blood flow due to vasodilation.^22,23 The current study used safe levels of hypoxia to study the TRBF and SO₂ changes. In humans, lower adenosine triphosphate (ATP) levels as well as release of the metabolite adenosine during hypoxia would lead to an increased retinal vessel diameter.⁶³ Few animal studies have reported other metabolic factors, such as retinal lactate,⁶⁴ adenosine, retinal relaxing factor,⁶⁵ and nitric oxide,⁶⁶ to mediate retinal blood flow response to hypoxia. 

In this study, retinal blood SO₂ is significantly reduced during changes in arterial O₂ tension, that is, below 100 mm Hg, above which there seems to be no significant change in retinal blood SaO₂, since the hemoglobin is almost 100% saturated. The increase in blood flow during hypoxia as shown in the present study compensates for the reduced retinal blood SO₂; thereby, demonstrating the regulation of inner retinal tissue during hypoxic environment. Our results demonstrated an inverse linear relationship between TRBF and retinal blood SO₂ in response to hypoxia as shown in Figures 6A and 6B. This finding is consistent with other cerebral^67,68 and retinal²² studies in literature. 

At the same time, decreased blood flow during hyperoxia is due to the vasoconstricting ability of retinal vessels. The higher concentration of O₂ in retina leads to an increased retinal arteriolar and venular blood SO₂. Also, the dissolved oxygen from choroid might as well increase the O₂ concentration in the retina considerably.^69,70 

There are possible limitations involved with this study. The TRBF derived using DOCTORC software needs several manual input in terms of grading, such as defining the cross-sectional area for retinal vessels from the Doppler OCT image and assigning confidence score based on the Doppler signal. The Doppler signal achieved was not uniform among all the scans from the same subject under various P_ETO₂ levels; this might have underestimated or overestimated the blood flow and vessel area estimation. Due to the long study duration (∼3 hours), subject's lack of concentration to fixate, and eye movements, image registration limitations using a manual system might have possibly influenced the quality of scans obtained or the quality of the acquired data. In Figure 1, the implausible changes in saturation along some of the vessels are due to artifact secondary to imperfections in image registration. This possibly could have attributed to the variability in the results. Keeping in mind that the image acquisition and image analysis were performed by trained personnel, the influence of human error could be considered minimal. Retinal SO₂ was measured in a single superior or inferior temporal retinal arteriole and venule close to the optic nerve head. This approach, rather than measuring total retinal SO₂, was undertaken for a number of reasons. There is a known marked regional variation in retinal SO₂ between the hemifields.⁷¹ The use of summary statistics to describe “total” retinal SO₂ values will result in the loss of the technique to identify localized change. To our knowledge, this is the first study to report the TRBF and retinal blood SO₂ measurements simultaneously in healthy subjects under conditions of hypoxia and hyperoxia. 

In conclusion, Doppler SD-OCT and MHRC could provide reliable and reproducible TRBF and retinal blood SO₂ measurements, respectively. By using a novel gas provocation technique to manipulate safe levels of P_ETO₂, we have demonstrated that both techniques could detect changes and showed the anticipated physiological response. In other words, increase in arterial P_ETO₂ from baseline decreases the TRBF. Conversely, decreasing the arterial P_ETO₂ from baseline increases the TRBF with simultaneous reduction in the retinal blood SO₂. Retinal blood flow and SO₂ measurements performed under safe levels of hypoxia and hyperoxia were repeatable in healthy adults. 

Acknowledgments

Supported by the Ontario Research Fund - Research Excellence Award, Vision Science Research Program, University of Toronto, and an anonymous donor. 

Disclosure: K. Rose, None; S.I. Kulasekara, None; C. Hudson, Optovue, Inc. (F), Optina Diagnostics (F), P