Impaired tissue blood supply is an important factor in the pathogenesis of retinal diseases.1–3 Therefore, quantification of ocular blood flow is valuable in assessing and treating individuals with retinal diseases. The Retinal Function Imager (RFI) (Optical Imaging Ltd., Rehovot, Israel)4–8 is a new non-invasive, multi-parameter functional imaging instrument. This instrument is based on differential imaging that directly detects moving red blood cells in retinal vessels and provides blood flow velocity measurements in secondary and tertiary branches of arteries and veins.7,9
Variability is a major concern associated with all techniques that evaluate ocular blood flow and limits their use in clinical practice and clinical trials.10 In addition to the variability of the instrument, physiological variability of retinal blood flow exists even in healthy individuals.11–13 Systemic pathological states such as hypertension are associated with structural microvascular abnormalities in the retina.14–16 Advancing age and increased blood pressure are associated with decrease in retinal vessel diameter, more in arteries than in veins.13,16–19 There are limited data on the effects of these factors on functional retinal parameters such as blood flow velocity.
Aging causes widespread physiological decline that reduces functional capacity and increases susceptibility to disease.20,21 Specifically, a high frequency of ophthalmic pathologies such as cataract, glaucoma, and age-related macular degeneration are associated with advancing age.21 Ischemic-related microangiopathies of the retina are the most common causes of blindness in developed countries.1 Therefore, understanding the mechanisms underlying the pathophysiology of retinal microcirculation is of fundamental clinical importance. To understand the effect of pathologies on blood flow, it is necessary to first understand how blood flow is affected in healthy individuals.
This study’s objective was to verify the reproducibility of the retinal blood flow measurements by the RFI and to determine the effect of physiological factors on the retinal hemodynamic state in healthy retinas. The parameters evaluated were age, heart rate, systemic blood pressure, smoking habits, retinal thickness, sex, and interocular differences.
Patients and Methods
Patients were included if they had no signs of ophthalmic pathology or intraocular pressure (IOP) greater than 21 mm Hg. Candidates in whom the retina could not be seen clearly because of media opacity or a poorly dilating pupil and individuals who had undergone eye trauma or surgery during the 6 months prior to recruitment were excluded (except for uneventful cataract extraction). One hundred thirty-two eyes of 74 individuals with no ophthalmic pathology were enrolled in this study. After RFI imaging, 18 eyes were excluded due to insufficient image quality. The final study population included 114 eyes of 67 volunteers. Of 114 eyes included in the analysis, 100 had clear lenses, 12 had mild nuclear sclerosis cataract, and two had previous cataract extraction surgery. In both cases, the surgery was performed more than 6 months prior to RFI scanning.
The study was approved by the Institutional Review Board/Ethics Committee of the Tel Aviv Medical Center, and adhered to the tenets of the Declaration of Helsinki and U.S. Health Insurance Portability and Accountability Act. A written informed consent was obtained from all participants after explanation of the nature and possible consequences of the study.
All participants underwent a comprehensive ophthalmic evaluation that included a medical history, assessments of best-corrected visual acuity and manifest refraction, IOP measurement by Goldmann applanation tonometry, slit-lamp examination before and after pupillary dilation, retinal nerve fiber layer (RFNL) thickness assessment, total macular volume and central macular thickness using optical coherence tomography (Stratus OCT Model 3000; Carl Zeiss Meditec AG, Berlin, Germany), and three sets of RFI images of 20° centered on the fovea. Pupils were dilated with 1% tropicamide and 2.5% phenylephrine. When eligible, both eyes of a patient were included. All participants had blood pressure measurement subsequent to RFI imaging (digital automatic blood pressure monitor M4-I; OMRON Healthcare Europe BV, Hoofddorp, Netherlands). Mean arterial pressure (MAP) was calculated as diastolic blood pressure plus one-third (systolic blood pressure minus diastolic blood pressure). Mean ocular perfusion pressure (MOPP) was calculated as two-thirds MAP minus IOP.
We used the RFI5,8 to measure blood flow velocity and heart rate. Briefly, the RFI extended the functionality of fundus camera optics by the addition of a stroboscopic flash lamp system and a computerized digital camera. The system provided up to eight flashes of illumination at close, computer-controlled intervals (typically 17.5 msec between flashes). For the blood flow velocity operating mode, the illumination filter was a green (“red-free”) interference filter with transmission centered at 548 nm and a bandwidth of 17 nm. To control the effect of heart pulsation on flow velocity measurements, a probe was attached to the participant’s finger or earlobe, allowing image acquisition to be synchronized to a constant, selected phase of the patient’s cardiac cycle. The exposure intervals of the digital camera were synchronized to the flash discharge. The digital pictures were captured, stored, and processed by differential imaging7,9 that directly detected moving red blood cells in retinal vessels. The measured velocity in secondary and tertiary branches of arteries and veins was recorded by superimposing it on the fundus image (mm/sec; Fig. 1, video available at www.healio.com/osli). Vessel segments were assigned relative densitometric scores that reflect vessel width, allowing subdivision of vessel segments into two categories: narrow and wide.
Figure 1. Blood flow velocity imaged by the Retinal Function Imager (Optical Imaging Ltd., Rehovot, Israel) in secondary and tertiary branches of arteries (red) and veins (purple). The numbers represent velocity (mm/sec). Arteriole velocity values are presented with negative values, representing the direction of flow. Scale bar = 500 μm.
The RFI scan was repeated at least three times for every measured region. For each segment, the coefficient of variance (standard deviation/mean) of the measured velocity was calculated. Image quality was evaluated both subjectively and objectively. The investigators evaluated the optical resolution, light intensity, and focus of the images. Image series were excluded if visible flow was not detected in the apparent vessels. Segments were excluded if high coefficient of variance between series (> 45%) was detected. This typically occurred in segments that were near the edge of the imaged region, in areas that were unfocused or poorly illuminated, or in segments that were not the optimal length (100 to 150 μm). Additionally, the entire image series was excluded if more than 33% of the vessels were excluded.
To assess the variability of the RFI blood flow velocity measurement, we calculated the coefficient of variance of all segments between each repeated measurement. We used segments common to all series and calculated each series average. The standard deviation of the series’ values divided by their average resulted in the coefficient of variance. In addition, an inter-visit analysis of variability was performed on five eyes of five healthy individuals. These individuals were imaged on two separate visits, each time three series were acquired. The reliability of width assessment was also evaluated by comparing repeated measurements of 18 participants to calculate how many vessel segments were miscategorized between narrow and wide.
Results are expressed as mean ± standard deviation and the change in velocity per 10 units of the effectors. Statistical analysis was performed using SPSS software, version 16 (SPSS Inc., Chicago, IL). The characteristics of the study population were compared using mixed linear models, taking into account the repeated measures of each outcome in the two eyes tested because the data of some participants were taken from both eyes. Interocular differences were calculated using paired Student’s t test. Inter-variability was calculated using interclass correlation. A mixture of Gaussians was used in a non-linear regression model for producing the velocity as a function of heart rate and age, avoiding overfitting. Significance was set at a P level of less than .05.
Participants’ characteristics are described in Table 1. The average velocity measured by the RFI of all participants included in the study was 4.2 ± 0.9 mm/sec in the retinal arterioles and 3.3 ± 0.8 mm/sec in the retinal venules. Variability of the measurements from the RFI was assessed at several levels. The average intra-visit variability as assessed in a subgroup of 20 participants by coefficient of variance was 7.5% ± 3.7% (Table 2). For measurements from the same participant on different days (inter-visit variability), the average interclass correlation coefficient was r = 0.744. The miscategorization of segmental width occurred on average in 13.6% ± 7% of the segments.
Table 1: Demographics of the Study Group
Table 2: Variability Measurement Characteristics
Velocity measurements in the same participants (n = 46) were not different between right and left eyes (paired t test, P > .05). We found a good correlation between arterial velocity (Fig. 2; r = 0.69, P < .001) and venous velocity (r = 0.62, P < .001) in right and left eyes.
Figure 2. The relationship between average blood velocity in arterioles (black) and venules (gray; mm/sec) between the right and left eye of the same participant. Linear trend line in black for arterioles and gray for venules.
We found a positive correlation between arterial and venous blood flow velocity and heart rate (n = 114, Fig. 3). For heart rate increase of 10 beats per minute (bpm), there is an increase of 0.21 mm/sec (5%; 95% confidence interval, 0.03 to 0.39; P < .05) in arterial velocity and 0.22 mm/sec (6.67%; 95% confidence interval, 0.06 to 0.38; P < .01) in venous velocity, adjusted for age and MAP.
Figure 3. The relationship between average blood velocity in arterioles (black) and venules (gray; mm/sec) and heart rate (HR; beats per minute, bpm). Not adjusted for age and mean arterial pressure. Linear trend line in black for arterioles and gray for venules.
Systemic Blood Pressure
We found a positive correlation between MAP (n = 104) and arterial retinal blood flow velocity (Fig. 4; P < .01). Arterial velocity increased as MAP increased, adjusted for age and heart rate. A 0.25 mm/sec (5.95%, 95% confidence interval, 0.08 to 0.42) increase in arterial velocity is evident for each 10 mm Hg increase in MAP. Such a correlation was not evident with venous retinal blood flow velocity. As expected, there was also a positive correlation between MAP and the arterial-to-venous ratio, which increases 0.1 (95% confidence interval, 0.05 to 0.14; P < .01) for each 10 mm Hg increase in MAP, adjusted for age and heart rate.
Figure 4. The relationship between the average blood velocity in arterioles (mm/sec) to mean arterial pressure (MAP; mm Hg). Not adjusted for age and heart rate. A linear trend line in black.
IOP and MOPP
We did not find correlation of blood flow velocity to IOP (n = 82 eyes) following adjustment for heart rate, age, and MAP. However, MOPP (n = 77 eyes) was strongly correlated to arterial-to-venous blood velocity ratio (mean change of 0.05 mm/sec for each 10 mm Hg increase in MOPP; 95% confidence interval, 0.02 to 0.08; P < .01, adjusted for heart rate and age). Velocity in the venules and arteries alone did not correlate to MOPP.
In participants older than 40 years (range: 40 to 81 years; mean = 56.6 ± 10.6 years; n = 64), we found a negative correlation between age and velocity in the venules (Fig. 5). There is a decrease in velocity in the venules of 0.32 mm/sec (95% confidence interval, −0.55 to −0.09; P < .01; 9.7% decrease) per decade above 40 years, adjusted for heart rate and MAP. In participants younger than 40 years (range: 19 to 40 years; mean = 28.4 ± 6.1 years; n = 50), there was no significant effect of age on retinal blood flow velocity. When studying the velocity as a function of heart rate and age using a Gaussian model, we found that the starting age of decrease in velocity is a function of heart rate (ie, the decline in velocity is earlier than 40 years in individuals with higher heart rate and begins later than 40 years in individuals with lower heart rate).
Figure 5. The relationship between average blood velocity (mm/sec) in venules and age (years) in participants older than 40 years. Not adjusted for mean arterial pressure and heart rate. Linear trend line in black.
In our study group, there were 20 eyes of 14 past smokers, mainly from the older group (mean: 61.4 ± 14.2 years). We compared past smokers and non-smoking participants (87 eyes). We found no significant effect of past smoking on blood velocity following adjustment for MAP, heart rate, and age in mean velocity of all segments. However, a trend of decrease in blood velocity was evident in the large arterioles in past smokers compared to non-smoking participants (mean difference = −0.59 mm/sec; 95% confidence interval, −1.19 to −0.01; P = .05).
Our study group included 24 men and 44 women (n = 37 and 77 eyes; mean age = 53.1 ± 14.4 and 40.0 ± 16.0 years, respectively). Velocity in wide venules was higher in women compared to men (mean difference = 0.39 mm/sec, 11.82%; 95% confidence interval, 0.74 to 0.03; P < .05) even adjusting for age, heart rate, and MAP. The large arterioles showed a similar trend, but this was not significant (mean change = 0.42 mm/sec; 95% confidence interval, 0.88 to 0.04; P = .075). In men and women older than 45 years, we found no difference in the velocity in all sizes of vessels with and without adjusting for heart rate and MAP.
Retinal Thickness and Volume
No correlation was found between blood velocity and peripapillary RNFL thickness (n = 89 eyes), total macular volume, or central macular thickness (n = 83 eyes) as measured by optical coherence tomography after adjusting for heart rate, MAP, and age.
We investigated the possible effects of different physiological factors on retinal blood flow velocity. We did not detect interocular differences in velocity between eyes of the same participants. Retinal arterial velocity correlated with increased heart rate (5% velocity increase per 10 bpm heart rate increase) and MAP (5.95% velocity increase per 10 mm Hg MAP increase). Increase in venous velocity correlated only with heart rate and increased by 6.67% per 10 bpm. A reduction of venous velocity was observed with increasing age, decreasing by 9.7% per decade. Velocity in large venules was 11.82% higher in women compared to men. We found no effect of hypertension on blood velocity or correlation with RNFL thickness.
Coefficient of variance estimation produced high correlations between velocity measurements; hence, the RFI measurements of retinal velocity were reproducible. These findings were comparable to the coefficient of variance of velocity measurements using scanning laser Doppler flowmetry (10% ± 3.4%), blue field entoptic technique, color Doppler imaging (9.3% ± 3.4%)10 and flow measurements using Doppler Fourier-domain optical coherence tomography (10.9%).22
Average velocities of 4.2 mm/sec in arterioles and 3.3 mm/sec in venules were measured by the RFI. Typical values using other instruments are presented in Table 3. Measurements by color Doppler imaging, laser blood flowmeter, and laser Doppler velicometry are taken from vessels larger than the secondary and tertiary branches of vessels that the RFI is measuring. In terms of vessel location, the RFI is mainly comparable to the fluorescein angiography and blue field entoptic techniques. Yet, one must bear in mind that the velocity of dye flow and leukocytes that are measured by these methods may be different from red blood cell flow.
Table 3: Typical Values of Blood Flow Velocity Using Different Techniques
To compare blood flow in a diseased eye to that in a fellow eye, standard data from two healthy eyes of participants are required. We found no interocular differences between right and left eyes, consistent with previous studies.25–29 Consequently, we used both eyes for our study. This interocular equivalency facilitates future studies comparing a healthy eye to a diseased fellow eye.
We found a positive correlation between arterial blood flow velocity and MAP as formerly found by measuring retinal white blood cell flux with the blue field entopic technique12 and using laser Doppler velocimetry.30 The effect is more pronounced in the arteries because the blood pressure gradient is steeper in the arteries.31 Significant correlations between blood pressure and flow velocity in the middle cerebral artery have been previously shown in healthy individuals.32 We did not find a correlation of retinal blood flow velocity with normal range IOP as maintained in steady state. Other studies that induced acute elevation of the IOP found changes in the retinal blood flow velocity.1 Naturally, the MAP and the MOPP, which is derived from the IOP and MAP, were associated with the arterial-to-venous ratio.
After the age of 40 years, increased age was associated with a significant reduction of retinal venous velocity. Similar findings were found with other instruments.25,27,33 A decrease with age in retinal arteriolar and venous diameter17–19 and blood flow was also reported.11,20,33,34 Loss of retinal cells, such as ganglion cells, photoreceptors, and retinal pigment epithelial cells, with age27 could reduce ocular perfusion demand, thus reducing blood flow. A decrease in blood flow velocity with age is also consistent with previous findings showing increases in the diameter of the foveal avascular zone, suggesting closure of capillaries with increasing age.27
Aging also induces loss of arterioles via reductions in growth hormone and insulin-like growth factors.21 The remaining vessels have reduced distensibility due to compromise of endothelium-dependent vascular relaxant mechanisms.21 There is an increased incidence of atherosclerosis with age that might also affect vessel function.11,33 We found that the age from which velocity starts to decline is related to heart rate. Increased heart rate may induce earlier exhaustion of the vessel wall, thus causing earlier decline of velocity as a function of age, or higher heart rate may be a compensation to lower velocities that declines with age.
We found similar velocities in a small group of past smokers compared to the non-smoking participants after adjusting for age, heart rate, and MAP. Nevertheless, in wide arterioles a trend of reduced velocity was found in past smokers compared to non-smoking participants that did not reach significance. A decreased ophthalmic artery peak systolic velocity in smokers and past smokers was previously found using color Doppler imaging.26 Additionally, smoking has systemic effects on circulation such as on heart rate and MAP. Thus, although studies exploring vessel diameter found differences between non-smokers and past smokers,13 these may be compensated for by the altered heart rate and blood pressure or other hemodynamic effectors to normalize velocity.
Cerebral blood flow in women is higher than in men of all ages.35 Thus, it is not surprising that we found higher retinal blood flow velocity in women compared to men. Women have larger artery-to-vein diameter ratios, resulting from wider arteries.18 The gender difference may be explained by the effect of estrogen and other hormones, which are known to be vasodilators.36 We did not perform a comparison between young males and premenopausal females due to the low number of young males in our study. However, comparison of women and men older than 45 years revealed no difference between them, implying the difference in velocity found between men and women may be attributed to the young premenopausal females.
We found no correlation between peripapillary RNFL thickness or central macular thickness by optical coherence tomography and blood flow velocity. A previous study showed similar lack of correlation between peripapillary RNFL thickness and perfusion parameters in healthy indviduals.37 However, significant correlation was found between venous retinal blood flow velocity and central retinal volume using the RFI in participants with retinal pathologies.38
Some study limitations should be noted. First, due to the cross-sectional design of the study, information about the effect of age was collected from different individuals at different ages and not from the same individuals over time. Second, there may be a sampling bias toward “normal” retina with age because older adults had a higher prevalence of retinal pathologies than individuals who were excluded from this study. Third, interpretation of the results of any retinal blood flow analysis must be performed with care because of the complex interactions of hemodynamic parameters that may be influenced by other unmeasured factors (eg, physical activity, excitement, and caffeine consumption) or unknown factors (eg, genetic factors and inflammatory markers). Finally, the RFI width measurement is not sensitive enough to extract flow volume data from velocity measurements.
The RFI proved to be a reproducible tool for assessing retinal blood flow velocity. The unique feature of this instrument is its ability to study the functional hemodynamic status in a large range of arterioles and venules simultaneously. Age, heart rate, and MAP influence the velocity measured and should be accounted for when assessing velocity under pathological conditions.