Conjunctival Vessels in Diabetes Using Functional Slit Lamp Biomicroscopy

Abstract

Purpose:

This study used functional slit lamp biomicroscopy (FSLB) to quantify conjunctival microvessel parameters in individuals with and without diabetes and examined whether these metrics could be used as surrogate markers of diabetes-related complications.

Methods:

A cross-sectional study of 98 controls (C), 13 individuals with diabetes without complications (D–C), and 21 with diabetes and related complications (D+C), which included retinopathy, nephropathy, neuropathy, and cardiovascular-, peripheral vascular-, and cerebrovascular diseases, was performed. Bulbar conjunctival metrics (venule diameter, length, axial velocity [Va], cross-sectional velocity [Vs], flow [Q], and branching complexity) were measured using FSLB (digital camera mounted on traditional slit lamp).

Results:

The mean age was 60 6 11 years, and demographics were similar across the groups. Va and Vs significantly differed between groups. Va was 0.51 ± 0.17 mm/s, 0.62 ± 0.17 mm/s, and 0.45 ± 0.17 mm/s in the C, D–C, and D+C groups, respectively (P = 0.025). Similarly, Vs was 0.35 ± 01.12, 0.43 ± 0.13, and 0.32 ± 0.13 mm/s in the C, D–C, and D+C groups, respectively (P = 0.031). Black individuals had increased Va, Vs, and Q compared with White individuals (P < 0.05), but differences in velocities persisted after accounting for race. Among patients with diabetes, Va and Vs correlated with number of organ systems affected (Va: ρ = −0.42, P = 0.016; Vs: ρ = −0.41, P = 0.021). Va, Vs, and Q significantly (P ≤ 0.005) discriminated between diabetic patients with and without complications (area under the receiver operating curve for Va = 0.81, Vs = 0.79, Q = 0.81).

Conclusions:

Bulbar conjunctival blood flow metrics measured by FSLB differed between controls, diabetic patients without complications, and diabetic patients with complications. FSLB is a quick, easily accessible, and noninvasive alternative that might estimate the burden of vascular complications in diabetes.

More than 30 million adults in the United States were estimated to have diabetes mellitus in 2015, according to the Centers for Disease Control and Prevention.¹ This number amounted to more than 12% of the adult population, increasing with age and extending to more than 25% among those aged 65 years and older. It is a leading cause of early illness and death, and prevalence continues to rise in the United States and around the world.² Complications of diabetes manifest as microvascular (retinopathy, nephropathy, and neuropathy) and macrovascular disease (atherosclerosis leading to cardiovascular, peripheral vascular, and cerebrovascular diseases). The burden on patients and the health care system from diabetes and related complications is considerable. In 2014, there were 7.2 million hospitalizations of US adults with diabetes included in their discharge diagnoses.¹ The total direct and indirect cost of diagnosed diabetes in 2012 was estimated to be $245 billion,³ with more than half of costs related to complications such as myocardial infarction, stroke, end-stage renal disease, retinopathy, and foot ulcers.⁴

Although associated with high morbidity, diabetes-related complications can be prevented with glycemic management.^5,6 In addition, the progression of diabetes-related complications can be slowed by interventions such as blood pressure and lipid control, angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers for kidney disease, and laser photocoagulation for retinopathy.⁷ As such, patients with diabetes require ongoing evaluation for screening and monitoring of complications. Diabetic complications are especially difficult to detect because symptoms might not present until significant pathological damage has occurred. Routine examinations are extensive and include blood and urine tests and eye and foot examinations. Such frequent and invasive evaluations can be cumbersome to patients,⁸ providers, and the health care system.¹ Thus, there is a need for improved detection of diabetic end-organ damage.

A novel approach for evaluating vascular status in diabetes is in vivo imaging of bulbar conjunctival hemodynamics, given the accessible location of conjunctival vessels. In vivo imaging of conjunctival vasculature has been achieved through several different methods including modified scanning laser ophthalmoscopy, adapted slit lamp biomicroscopy digital imaging, slit lamp stereomicroscopy, laser Doppler flowmetry, orthogonal polarized spectral imaging, computer-assisted intravital microscopy, and anterior segment optical coherence tomography angiography. Using these approaches, several studies have found correlations between diabetes and conjunctival vessel parameters. These include increased mean vessel width,^9–12 lower arteriole–venule ratios,^13,14 decreased macrovascular (vessels > 40 μm in width) tortuosity,¹⁵ increased microvascular (capillary-sized vessels < 25 μm in width) tortuosity,^14–16 higher density,¹⁷ abnormal microvascular distribution,^9,11 slower velocity and flow,^9,11,14 and overall abnormal indices^11,18 in diabetic patients when compared with controls.

Functional slit lamp biomicroscopy (FSLB) was developed to expand on the above-mentioned technologies. FSLB is composed of a traditional slit lamp with a commercially available digital camera capable of capturing advanced quality imaging at extremely high magnifications and imaging speeds. Similar to the other systems, it can measure conjunctival microvessel diameter, blood flow velocity, and blood flow rate in real-time. In addition, it can image both small and large fields and construct noninvasive microvascular perfusion maps (nMPMs) to quantify vessel branching complexity. This added capability can increase sensitivity for detecting earlier vascular changes.¹⁹ FSLB is noninvasive and quick, taking approximately 5 minutes per patient to set up, obtain imaging, and check the quality of images on site. It has been validated for analysis of both hemodynamics and branching complexity of conjunctival vessels.^20–22 It has demonstrated the ability to detect changes in conjunctival microvasculature, such as that which occurs with the placement of a contact lens.²⁰ In addition, we have previously shown that conjunctival vessel metrics measured by FSLB can serve as a predictor of cardiovascular mortality, based on correlations between blood flow velocity and rate and the Framingham risk score.²³

Based on its abilities, we aimed to expand on the above-mentioned studies and use FSLB to evaluate conjunctival vasculature in individuals with varying severity of diabetes, namely those with and without diagnosed vascular complications, when compared with a control population. We hypothesized that individuals with diabetes and vascular complications would have altered conjunctival vessels metrics when compared to diabetic patients without vascular complications and controls. A quick, noninvasive, and reliable screening test could improve early detection of diabetic complications and translate into better patient and provider experiences.

MATERIALS AND METHODS

Study Design, Setting, and Population

This was a cross-sectional study of individuals prospectively recruited from the Miami Veterans Affairs (VA) Health care System Eye Clinic. This study was approved by the Miami VA Institutional Review Board, was conducted in accordance with the principles of the Declaration of Helsinki, and complied with the requirements of the US Health Insurance Portability and Accountability Act. Informed consent was obtained from the subjects after explanation of the nature and possible consequences of the study. Individuals underwent complete ocular surface examinations and were included in the study if they had otherwise healthy eyelid and corneal anatomy. Individuals were divided into 3 groups: diabetic patients with vascular complications (D+C), diabetic patients without complications (D–C), and healthy controls (C). We excluded individuals whose ocular or systemic disease might impact conjunctival vessel anatomy including individuals who wore contact lenses, used ocular medications with the exception of artificial tears, had ocular comorbidities (eg, pterygium, glaucoma, and infection), underwent cataract surgery within the past 6 months, had refractive, glaucoma, or retinal surgery, or had human immunodeficiency virus, sarcoidosis, graft-vs-host disease, or a collagen vascular disease.

Diabetic Risk Estimation

Medical history related to diabetes mellitus was obtained through chart review. For individuals with a diagnosis of diabetes, additional information was collected including duration of disease, most recent hemoglobin A1C (HbA1C), and history of diabetes-related complications such as retinopathy, nephropathy, neuropathy, and cardiovascular, peripheral vascular, and cerebrovascular diseases. Retinopathy was considered present if specified as diabetic retinopathy or diabetic macular edema. Nephropathy was considered present by early signs (positive urine albuminuria, either a spot urine with albumin-to-creatinine ratio ≥30 mg/g or a positive urine dipstick) or late signs (diagnosis of chronic kidney disease). Neuropathy was considered present if specified as diabetic neuropathy. Cardiovascular disease was considered present if the subject carried any of the following diagnoses: coronary artery disease, cardiomyopathy, or congestive heart failure. Peripheral vascular disease included peripheral arterial disease and foot ulcers, and cerebrovascular disease included strokes and transient ischemic attacks.

Demographics and Comorbidities

Demographic information such as age, sex, ethnicity, race, and smoking status was obtained. Ethnicity and race were categorized according to the US Census Bureau. In addition, medications and comorbidities such as hypertension and hypercholesterolemia were recorded.

Functional Slit Lamp Biomicroscopy

The FSLB imaging system is composed of a digital camera mounted on a traditional slit lamp, as described and validated in previous studies.^20–25 The digital camera used in this study was a commercially available Canon 60D model with Movie Crop Function capable of about ×7 magnification with high-speed video recording at 60 frames per second and no loss of image quality. Combined with the slit lamp potential of ×25 magnification, the FSLB system can reach about ×175 magnification. Green light was used to enhance resolution because red blood cells absorb more green light than surrounding tissues. FSLB captured the movement of red blood cell clusters in bulbar conjunctival vessels to measure blood flow velocity, flow rate, and vessel diameter.

Images and videos were taken of the temporal bulbar conjunctival vessels in the right eye. Individuals were asked to focus on a point in the far-left field of vision. To capture images, a slit lamp was placed 20 degrees nasally and set to half brightness at maximum size and magnification of ×16 to produce a field view of about 15.70 × 10.47 mm² (Fig. 1A). Videos were captured using a slit lamp placed 45 degrees temporally, set to maximum brightness with a magnification of ×25 and reduced to a field size of about 1.22 × 0.91 mm² (Fig. 1B). This yielded more than the suggested minimum of 1 pixel for 1 red blood cell to obtain an adequate signal-to-noise ratio.²⁶ At least 6 separate fields of view of temporal bulbar conjunctival vessels were obtained by video, each field lasting a few seconds. This ensured that more than 15 venules were captured for each patient because a previous study found that a sample of 15 venules yielded an acceptable standard error of 15%.²⁵

FIGURE 1.

Images of temporal bulbar conjunctival vessels using FSLB. A, Slit lamp magnification of ×16. Used to image nMPMs for fractal analysis. B, Still image from a video capturing movement of red blood cells, with total magnification of ×175 when combined with slit lamp magnification. Used to image hemodynamics and calculate mean vessel diameter, length, axial blood flow velocity, cross-sectional blood flow velocity, and flow rate.

Imaging Bulbar Conjunctival Hemodynamics

As detailed in previous studies,^20,25 custom software was used to semiautomatically process the videos and generate mean venule diameters (mm), lengths (mm), axial velocities (Va, mm/ s), cross-sectional velocities (Vs, mm/s), and blood flow rates (Q, pL/s). Only venules were processed in this study because pulse can affect blood flow in the conjunctival precapillary arterioles.²⁷ Based on their diameter (bulbar conjunctival venules are larger) and direction of flow (venules collect flow from bifurcation branches), venules were differentiated from arterioles. Va was measured by the space-time image technique, which was then used to calculate Vs and Q based on an equation for blood flow velocity in vessels with diameters less than 20 μm.²⁸

Imaging Bulbar Conjunctival nMPMs and Fractal Analysis

Similar to a procedure used to generate noninvasive capillary perfusion maps of the retina,²⁹ this study used a single image method to generate nMPMs of the bulbar conjunctiva, as described in a previous study.²⁰ Custom software was then used to analyze complexity of the microvascular morphology, represented as the monofractal (Dbox) and multifractal dimensions (D0). Dbox and D0 quantified the complexity of the branching pattern of the vascular networks, with D0 providing a more sensitive method to detect differences in the microvasculature than Dbox.³⁰

Statistical Analysis

Data were entered into a standardized database, and statistical analyses were performed using SPSS V.26.0 statistical package. Descriptive statistics were used to describe the study population. One-way analysis of variance tests were used to compare continuous variables, with post hoc least significant differences for pairwise comparisons. The χ² tests were used to compare categorical variables. Correlation coefficients were calculated to compare 2 continuous variables. Levene test and examination of residuals assessed variance heterogeneity and other model fit characteristics. Forward stepwise regressions investigated for effects of specific diabetes-related complications on conjunctival metrics. In individuals with diabetes, forward stepwise regressions were also fit to analyze the ability to discriminate between those with and without complications using conjunctival microvascular metrics and demographics as independent variables. Receiver operating characteristic curves were generated to evaluate conjunctival microvascular metrics as possible predictors of diabetes-related complications. A power calculation using our sample size of 136 and α of 0.05 yielded 71% power to detect a medium effect size and greater than 99% to detect a large effect size between the groups.

RESULTS

Demographics and Comorbidities

A total of 136 individuals were recruited from the Miami VA Health care System for this study. Baseline characteristics are depicted in Table 1. The mean age was 60 years with SD of 11 years. Most individuals were identified as men (118 [87%]), non-Hispanic (101 [74%]), and Black (82 [60%]). Thirty-eight (28%) individuals had a diagnosis of diabetes mellitus. Of the 38 individuals with diabetes, 4 were excluded from analyses of diabetes-related complications because of missing information from their charts regarding complications. This yielded 98 healthy controls (C), 13 with diabetes but no complications (D–C), and 21 individuals with diabetes-related complications (D+C).

TABLE 1.

Demographics

	All	Without Diabetes (C)	With Diabetes but No Related Complications (D−C)	With Diabetes and Related Complications (D+C)	P
N	136	98	13*	21*	—
Age, yrs, mean (SD)	59.84 (10.52)	59.62 (11.14)	62.08 (8.49)	59.62 (8.43)	0.728
Sex, male, n (%)	118 (87)	82 (84)	12 (92)	21 (100)	0.108
Ethnicity, Hispanic, n (%)	35 (26)	22 (22)	5 (39)	7 (33)	0.318
Race, n (%)					0.381
White	54 (40)	42 (43)	4 (31)	6 (29)
Black	82 (60)	56 (57)	9 (69)	15 (71)
Other	0 (0)	0 (0)	0 (0)	0 (0)
Smoking status, n (%)					0.609
Never	21 (15)	18 (18)	2 (15)	1 (5)
Past	63 (46)	43 (44)	5 (38)	11 (52)
Current	52 (38)	37 (38)	6 (46)	9 (43)
Hypertension, n (%)	86 (63)	54 (55)	10 (77)	20 (95)	0.001
Hypercholesterolemia, n (%)	75 (55)	48 (49)	10 (77)	15 (71)	0.044
HbA1C, %, mean (SD)	6.00 (1.29)	5.48 (0.55)	6.69 (0.60)	7.81 (2.03)	>0.001
Duration of diabetes, yrs, mean (SD)	—	—	4.69 (5.93)	5.86 (4.45)	0.519

Numerical variables (age, HbA1C, and duration of diabetes) were compared using 1-way analysis of variance tests. Categorical variables (sex, ethnicity, race, smoking status, hypertension, and hypercholesterolemia) were compared using Pearson χ² tests. Bolded P values reflect statistical significance (P > 0.05).

^*Four individuals with diabetes were excluded from analyses involving diabetes-related complications because presence or absence of diabetes-related complications could not be determined.

Between the 3 groups, there were no significant differences in age, sex, ethnicity, race, and smoking status. As expected, there were increased percentages of individuals diagnosed with hypertension (C: 55%, D–C: 77%, D+C: 95%; P = 0.001) and hypercholesterolemia (C: 49%, D–C: 77%, D+C: 71%; P = 0.044) in the diabetic groups compared with controls. HbA1C also reflected increasing levels with higher clinical severity (C: 5.48, D–C: 6.69, D+C; P < 0.001).

No significant differences in conjunctival microvascular metrics were found based on age, sex, smoking status, or diagnosis of hypertension or hypercholesterolemia. However, race affected conjunctival velocities, flow rate, and complexity of morphology. Individuals who were identified as Black had higher Va, Vs, Q, Dbox, and D0 compared with individuals identified as White, as depicted in Table 2. Levene test and examination of residuals did not reveal any problems with variance heterogeneity or other model-fit characteristics.

TABLE 2.

Conjunctival Microvascular Metrics Based on Race

	All	White	Black	P
N	136	54	82	—
Diameter, μm	21.44 (3.13)	21.09 (3.36)	21.68 (2.97)	0.291
Length, μm	245.07 (65.72)	235.50 (54.52)	251.38 (71.78)	0.169
Va, mm/s	0.51 (0.18)	0.45 (0.14)	0.54 (0.19)	0.003
Vs, mm/s	0.36 (0.12)	0.32 (0.09)	0.38 (0.13)	0.004
Q, pL/s	143.66 (69.27)	125.24 (55.51)	155.79 (74.87)	0.011
Monofractal dimension, Dbox	1.63 (0.06)	1.61 (0.06)	1.64 (0.05)	0.001
Multifractal dimension, D0	1.66 (0.05)	1.65 (0.05)	1.67 (0.05)	0.006

Values are given as mean (SD). Comparison was performed using 1-way analysis of variance tests. Categories determined by the US Census Bureau. No individuals identified as another race (Asian, American Indian and Alaskan Native, Native Hawaiian and other Pacific Islander, or other). Bolded P values reflect statistical significance (P < 0.05).

Diabetes Status

There were no significant correlations between conjunctival microvascular metrics and HbA1C or duration of diabetic disease. However, comparisons of conjunctival metrics between the C, D–C, and D+C groups revealed significant differences in velocities, as depicted in Table 3 and graphically in Figure 2. Post hoc pairwise comparisons revealed that individuals in the D–C group (Va = 0.62 ± 0.17 mm/s; Vs = 0.43 ± 0.13 mm/s) had significantly higher Va and Vs compared with those in the C group (Va = 0.51 ± 0.17 mm/s, P = 0.032; Vs = 0.35 ± 0.12 mm/s, P = 0.029) and in the D+C group (Va = 0.45 ± 0.17 mm/s, P = 0.007; Vs = 0.32 ± 0.13 mm/s, P = 0.009). Although the mean velocities of the patients in the D+C group were lower than those of controls, these differences were not found to be statistically significant in our study. The Q values followed a similar pattern but did not reach statistical significance.

TABLE 3.

Conjunctival Microvascular Metrics Based on Diabetes Status

	All	Without Diabetes (C)	With Diabetes but No Related Complications (D−C)	With Diabetes and Related Complications (D+C)	P
N	136	98	13*	21*	—
Diameter, μm	21.44 (3.13)	21.58 (3.05)	21.36 (2.09)	21.03 (3.96)	0.758
Length, μm	245.07 (65.72)	250.42 (69.89)	240.78 (51.60)	229.45 (55.30)	0.404
Va, mm/s	0.51 (0.18)	0.51 (0.17)	0.62 (0.17)	0.45 (0.17)	0.025
Vs, mm/s	0.36 (0.12)	0.35 (0.12)	0.43 (0.13)	0.32 (0.13)	0.031
Q, pL/s	143.66 (69.27)	146.01 (73.90)	170.75 (40.89)	121.66 (54.39)	0.123
Monofractal dimension, Dbox	1.63 (0.06)	1.63 (0.06)	1.63 (0.04)	1.62 (0.07)	0.804
Multifractal dimension, D0	1.66 (0.05)	1.66 (0.05)	1.67 (0.39)	1.66 (0.07)	0.803

Values are given as mean (SD). Comparison was performed using 1-way analysis of variance tests. Bolded P values reflect statistical significance (P < 0.05).

^*Four individuals with diabetes were excluded from analyses involving diabetes-related complications because presence or absence of diabetes-related complications could not be determined.

FIGURE 2.

Va, Vs, and flow rate (Q) based on diabetes status. Box-and-whisker plots with outliers excluded, X marks mean. Vascular parameters (Va, Vs, and Q) were compared using 1-way analysis of variance tests between diabetes status groups (C, D–C, and D+C). Va and Vs were statistically significantly different across groups (P = 0.025 and P = 0.031, respectively). Q did not demonstrate statistically significant differences (P = 0.123) but displayed similar patterns across groups. Post hoc pairwise comparisons for Va and Vs are displayed beneath the plots (*C vs. D–C, †D–C vs. D+C, ‡C vs. D+C). Post hoc pairwise comparisons were not performed for Q.

Although race and diabetes status were significantly associated with Va and Vs, no interactions were found between the 2 independent variables. Black individuals consistently demonstrated higher Va and Vs than White individuals across all diabetes status groups. The differences in Va and Vs across diabetes status persisted after accounting for the effect of race, with the D–C group having higher velocities (Va = 0.60 ± 0.05 mm/s; Vs = 0.42 ± 0.03 mm/s) than either the C group (Va= 0.50 ± 0.02 mm/s, P = 0.046; Vs = 0.35 ± 0.01 mm/s, P = 0.042) or the D+C group (Va = 0.43 ± 0.04 mm/s, P = 0.005; Vs = 0.31 ± 0.03 mm/s, P = 0.007). The D+C group demonstrated a trend toward lower Va compared with that of the C group. Race was also significantly associated with Q values, but diabetes status was not.

Diabetes-Related Complications

Among individuals with diabetes-related complications (n = 21), 13 had nephropathy (62%), 11 had cardiovascular disease (52%), 8 had neuropathy (38%), 7 had retinopathy (33%), 4 had cerebrovascular disease (19%), and 2 had peripheral vascular disease (10%). Blood flow metrics were not different when comparing individuals with and without a specific complication. However, blood flow metrics were inversely correlated with number of diabetes-related complications (Va: ρ = −0.421, P = 0.016; Vs: ρ = −0.406, P = 0.021) (Table 4).

TABLE 4.

Associations Between Conjunctival Microvascular Metrics and Number of Diabetes-Related Complications

	ρ	P
Diameter, μm	0.080	0.664
Length, μm	−0.068	0.713
Va, mm/s	−0.421	0.016
Vs, mm/s	−0.406	0.021
Q, pL/s	−0.343	0.055
Monofractal dimension, Dbox	−0.222	0.255
Multifractal dimension, D0	−0.103	0.603

Bolded P values reflect statistical significance (P < 0.05).

ρ = Spearman rho (nonparametric correlations).

All 3 hemodynamic parameters (Va, Vs, and Q) were found to significantly discriminate between diabetic patients with and without complications. As depicted in Figure 3, the area under the receiver operating curve was 0.81 for Va (95% confidence interval [CI] 0.66–0.95, P = 0.003), 0.79 for Vs (95% CI 0.64–0.94, P = 0.005), and 0.81 for Q (95% CI 0.65–0.96, P = 0.003).

FIGURE 3.

Receiver operating curves (ROC) with conjunctival hemodynamic parameters as predictors of risk for systemic complications in individuals with diabetes.

To assess for potential confounders, we reran the model including demographics (age, sex, ethnicity, and race), smoking status, diabetic parameters (HbA1C and duration of diabetes), and comorbidities (hypertension and hypercholesterolemia) in addition to the 3 hemodynamic parameters (Va, Vs, and Q). In this model, all hemodynamic parameters remained significantly associated with presence of end-organ damage in diabetic patients, in addition to hypertension. Of interest, race did not remain significantly associated with diabetes-related complications.

DISCUSSION

This study demonstrated 4 important findings regarding bulbar conjunctival microvasculature as detected by FSLB. First, conjunctival metrics differed between races, with Black individuals demonstrating higher velocities and flow and complexity of morphology across all diabetes status groups when compared with White individuals. Second, when comparing diabetes status and adjusting for race, diabetic patients without complications had the highest velocities and diabetic patients with complications had the lowest. Third, among individuals with diabetes, the number of organ systems affected by diabetes negatively correlated with conjunctival microvascular velocities. Fourth, conjunctival hemodynamic metrics were able to distinguish between individuals with and without vascular complications, as demonstrated by the receiver operating characteristic curves.

Of interest, the blood velocity differences noted between our groups were observed despite no differences in overall vessel diameter. Our findings share both similarities and differences to previous studies that examined conjunctival vessels in diabetes. One group found that individuals with diabetes had increased conjunctival vessels diameters and slower velocities than controls.⁹ However, the study analyzed conjunctival vessels greater than 50 mm in width, which were larger than the 20 mm microvessels examined in our study, and it did not differentiate by presence of complications within diabetes. In fact, another group examined various conjunctival vessels in diabetes and found increased diameters in larger (>40 μm) but not smaller (<25 μm) vessels when compared with controls.¹⁰

More similar to our methodology, 1 group used a slit lamp biomicroscope to analyze conjunctival blood vessels in controls, diabetic patients without clinically visible retinopathy, diabetic patients with nonproliferative diabetic retinopathy, and diabetic patients with proliferative diabetic retinopathy.³¹ In their study, the nonproliferative diabetic retinopathy group demonstrated significantly higher venular velocity compared with all other groups, whereas the proliferative diabetic retinopathy group demonstrated significantly lower velocity. Although our study groupings differed, both found patterns of initial increase in venule velocity, followed by a decrease with advancing diabetes.

We hypothesize that endothelial dysfunction contributes to the observed velocity differences.³² Among other functions, the endothelium is responsible for vasodilation through nitric oxide (NO) production. In diabetes, there is decreased NO bioavailability due to impaired NO production by the endothelium and increased NO inactivation by reactive oxygen species. This impairment in endothelial function exists early in diabetes, likely starting at the prediabetic stage, and is central to the development of microvascular and macrovascular complications.³³ Our observed pattern of increased blood velocities in diabetes without vascular complications and slower blood velocities in those with more severe disease might indicate an initial compensatory phase, followed by a decompensated phase. In the compensatory phase, hypoxia perpetuated by a lack of vasodilatory ability is initially placated by increased blood velocity to maintain oxygen delivery to tissues. However, over time, elevated vascular tone induces permanent vascular remodeling including thickening of the lumen wall, which decreases the internal lumen diameter and increases vascular stiffness.³⁴ This permanent remodeling hinders the movement of blood and eventually overcomes compensatory mechanisms, leading to overall decreased blood velocity. This pathophysiology likely underlies the beginning of overt signs of diabetic vasculopathy. In support of this hypothesis, we found that, among diabetic patients with vascular complications, blood velocities were inversely proportional to number of organ systems affected, demonstrating increased vasculopathy with increased organ damage and diabetes severity.

Likely because of a complex array of biological and environmental factors, individuals of Black race have a higher risk for developing vascular diseases such as hypertension and diabetes mellitus.³⁵ A meta-analysis of data examining vascular function between races and ethnicities revealed that, compared with normotensive patients of White ethnicity, normotensive patients of Black ethnicity demonstrated increased reactivity to sympathetic stimulation, lower reactivity to vasodilators, and smaller lumen diameters, all leading to increased vascular tone.³⁶ Studies have shown that even mild elevation of vascular tone results in significantly increased peripheral vascular resistance, which is the hallmark of hypertension and a central contributor to diabetic vasculopathy. In support with these findings, our study demonstrated that Black individuals had increased velocities and flow through conjunctival microvessels compared with White individuals, perhaps demonstrating compensatory means to provide adequate blood delivery. In addition, Black individuals had increased complexity of vessel morphology compared with White individuals, which could also reflect compensatory structural changes driven by increased vascular resistance.

As with all studies, our findings must be examined under consideration of potential limitations. These include a cross-sectional design, focused patient population, and reliance on chart review for diagnoses. the pilot nature of this study was to first discriminate between diabetic patients with and without complications using conjunctival microvasculature measured by FSLB, severity of diabetes-related complications was not analyzed in this study. Differential blood flow patterns between healthy controls and diabetic patients with and without overt end-organ damage without examination of end-organ damage severity have been demonstrated in the hands and feet in previous studies.³⁷ We recognize, however, that future studies are needed to investigate conjunctival microvasculature in relation to specific complications and severity of complications.

However, despite these limitations, this study demonstrated that FSLB detected differences in conjunctival blood velocity in individuals with diabetes and related complications, diabetic patients without related complications, and healthy controls. Given the need for noninvasive tests to monitor vascular status, FSLB might serve as a tool to improve the detection of diabetes-related complications. Future prospective studies are needed to validate its use as a predictive tool of early and late diabetic complications in diverse populations. The former is important as early preclinical disease (ie, microalbuminuria) is a time point at which interventions are more amenable to changing the course of the disease.