Carotid Disease and Retinal Optical Coherence Tomography Angiography Parameters in Type 2 Diabetes: The Fremantle Diabetes Study Phase II

Macrovascular and microvascular complications frequently coexist in people with diabetes (1). In the specific case of a relationship between carotid arterial disease and diabetic retinopathy (DR), however, the available data are inconsistent (2–5). This may, in part, reflect methods of ascertainment of DR. Color photographs may provide a limited view and thus miss peripheral disease, while macroscopic changes evident on the images may not reflect more extensive undetected microangiopathy (6). A more sensitive method of retinal imaging could clarify the relationship between proximal carotid disease and downstream microvascular changes.

Optical coherence tomography angiography (OCTA) is modern noninvasive technology that allows direct visualization of microvascular blood flow in the retina and choroid (7). OCTA imaging is efficient, with most images acquired in ∼6 s (8). OCTA uses motion contrast imaging to construct a detailed map of retinal blood flow (8). It compares the decorrelation signal of sequential B-scans of the same cross section (8). OCTA variables have been associated with DR (7) and used to detect preclinical disease through earlier detection of retinal abnormalities (9,10). OCTA has been shown to be at least equivalent to fluorescein angiography in characterizing retinal pathologies (11,12) and, in one study, was superior in identification of areas of nonperfusion (13). The clinical application of OCTA covers many retinal vascular conditions other than DR including retinal venous occlusion, retinal arterial occlusion, uveitis, and age-related macular degeneration (14).

Given inconsistent associations between carotid disease and DR in past studies (2–5) and the emergence of OCTA as a sensitive method of assessing the retinal microvasculature, the aim of the current study was to assess whether markers of carotid disease assessed using conventional ultrasonography are associated with OCTA retinal microvascular parameters in people with type 2 diabetes.

The Fremantle Diabetes Study Phase II (FDS2) is a community-based, prospective, observational study that recruited 1,551 participants with clinically diagnosed type 2 diabetes between 2008 and 2011 from a zip code–defined urban community surrounding the port of Fremantle in the state of Western Australia. The characteristics of FDS2 participants and those identified but not recruited have previously been described (15). A random sample of 360 FDS2 participants with type 2 diabetes who had been assessed in detail at their last (year 6) face-to-face study visit was invited to participate in this study, which was conducted between May 2018 and May 2019. The study was approved by the South Metropolitan Health Service Human Research Ethics Committee (project number RGS0000000805), and all participants provided written informed consent.

Participants underwent a detailed assessment including medical history, physical examination, usual-care biochemical tests, dilated color fundus photography, fundus autofluorescence imaging, OCTA, and carotid ultrasonography, conducted in a single visit. A Body Shape Index (ABSI) was calculated from height, weight, and waist circumference (16). Biochemical tests were performed on fasting blood and first morning urine samples using standard automated methods in a single nationally accredited laboratory. A trained registered nurse (J.J.D.) carried out the physical assessment according to a standardized protocol. This included anthropometry, supine blood pressure, pulse, ophthalmic assessment, and carotid ultrasonography.

Visual acuity was measured in a well-lit room with an Early Treatment Diabetic Retinopathy Study (ETDRS) chart. The best-corrected visual acuity letter score was used in analysis. Refractive error was assessed with the Retinomax K-plus2 (Nikon, Tokyo, Japan). Axial length (AL) was assessed using the Carl Zeiss IOLMaster 500 (Zeiss, Oberkochen, Germany). After visual acuity was assessed, participants had their fundus dilated using 0.5% tropicamide solution for photography and OCTA. Single-field color fundus photographs and fundus autofluorescence imaging (field of view 45°) were performed by use of a nonmydriatic retinal camera (CR2-AF; Canon, Tokyo, Japan). Using the color photographs, an experienced ophthalmologist (F.K.C.) graded DR presence and severity according to the International Clinical Disease Severity Scale for DR (17).

OCTA was performed with the RTVue XR3 Avanti (Optovue, Fremont, CA). The 3 mm × 3 mm macula scans were analyzed. Participants with a scan quality <6 out of 10 were excluded from analysis. Microvascular parameters assessed with use of OCTA were foveal avascular zone (FAZ) area (mm²), vessel density in the 300-μm ring (FD-300) surrounding the FAZ (%), parafoveal density (%) of vessels in the deep capillary plexus and superficial capillary plexus, full retinal thickness (μm) of the parafovea (inner limiting membrane to retinal pigment epithelium layers), and flow area (mm²) in the choriocapillaris (Fig. 1). A list and definition of these parameters are shown in Supplementary Table 1. The measurement of the FAZ area was corrected according to the Linderman formula FAZ_corrected = FAZ_nominal (AL_S/AL_M)², where AL_S is the AL of the participant in mm and AL_M is the AL assumed by the manufacturer (23.95 mm in this case) (18).

Carotid ultrasounds were performed using the IU22 system with a 3.0–9.0 MHz linear transducer (Philips Healthcare, North Ryde, Australia). A trained registered nurse (J.J.D.) performed the imaging, and all results were reviewed by a trained sonographer (A.M.B.) with additional oversight from a neuroradiologist (B.T.D.). Intima-media thickness (IMT) was measured, and the presence and degree of stenosis were assessed. With antero-oblique insonation, the far-wall carotid IMT was visualized bilaterally within the carotid bifurcation and the common carotid artery (CCA) at least 1.0 cm proximally from the bifurcation. The optimized transducer depth (usually 3.5 cm) was adjusted to avoid slice thickness artifacts. The gain was adjusted so that the least dense arterial wall interface was visible. Three edge-to-edge measurements were taken of the far wall, both at the carotid bifurcation and the CCA, without the zoom function. The images were captured during systole at the R waves over three to four cardiac cycles. The mean of the three valid IMT measurements per site was used in analysis. The mean of IMT measurements was both assessed as a continuous variable and dichotomized as <1 mm and ≥1 mm as an established clinically significant cut point (19).

The presence of carotid plaque assessed as hard (echogenic and/or calcific), soft (hypoechoic, echogenic without calcification), or mixed was defined as focal wall thickening ≥50% greater than that of the surrounding vessel wall or as a focal region with IMT >1.5 mm protruding into the lumen and distinct from the adjacent boundary. Stenoses were defined as 1) 0% (normal waveform/image), 2) <15% diameter reduction assessed as deceleration spectral broadening and peak systolic velocity (PSV) <125 cm/s, 3) 16–49% diameter reduction with pansystolic spectral broadening (PSB) and PSV <125 cm/s, 4) 50–69% diameter reduction with PSB, PSV ≥125 cm/s, and end diastolic velocity (EDV) <110 cm/s or ratio of internal carotid artery PSV to CCA PSV >2, 5) 70–79% diameter reduction with PSB and at least one PSV >270 cm/s or EDV >110 cm/s or ratio of internal carotid artery PSV to CCA PSV >4, 6) 80–99% diameter reduction with criteria as in 5 plus EDV >140 cm/s, and 7) occluded with no flow and terminal thump.

Three sample size calculations were performed that assumed the prevalence of carotid disease in people with type 2 diabetes was 40% (2,20) and used a type 1 error probability of 0.05. The first concerned the ability to detect differences in the presence of DR, as there are limited published OCTA data. If the prevalence of DR in those with and without carotid disease was 30% and 12%, respectively (21,22), 184 participants would be needed to reject the null hypothesis with 80% power. For the other two calculations, 133 participants would be needed at ≥80% power to reject the null hypotheses for 1) FAZ with an estimated SD of 0.27 mm² and difference in means in those with and without carotid disease of 0.18 mm² and 2) the parafoveal vessel density in the deep capillary plexus with an SD of 0.1% and difference in means by carotid disease status of 0.05% (23).

Statistical analyses were conducted with the computer packages IBM SPSS for Windows (version 25.0; IBM, Armonk, NY) and Stata (version 15.1; StataCorp, College Station, TX). Data are presented as proportions, mean ± SD, geometric mean (SD range) or median (interquartile range [IQR]). The OCTA variables were split into quartiles for bivariable comparisons. Comparisons of two independent samples were by Fisher exact test or χ² test for categorical variables, one-way ANOVA for normally distributed continuous variables, and Kruskal-Wallis test followed by Dunn post hoc test for nonnormally distributed variables. Post hoc tests were performed to assess differences between quartiles. The Bonferroni correction for multiple comparisons was applied in all cases.

The generalized estimating equation extension of generalized linear models with an unstructured matrix was used to determine the association between carotid disease and OCTA parameters. With this method, one can adjust for the correlation between the two eyes in one person, enabling inclusion of both eyes in the analysis. Firstly, ipsilateral carotid disease variables were added to an unadjusted model. Secondly, associations of each OCTA parameter with P < 0.2 in bivariable analysis were used to generate the most parsimonious model for that OCTA variable. Carotid disease–related variables were then added one at a time to the most parsimonious model for assessment of whether they were independently associated. Thirdly, the correlation between left and right continuous carotid variables and OCTA variables was assessed with use of either Pearson correlation or Spearman rank correlation (r_s) and the difference between the categorical carotid variables with use of the McNemar test. For characterization of the relationship between carotid disease and OCTA parameters, contralateral carotid disease–related variables were also assessed in unadjusted and adjusted models. Since the mean IMT measurements of the CCA and bifurcation were right skewed, natural logarithm (ln) transformations of these variables were used in analysis.

We recruited 273 participants of the 360 invited FDS2 subjects (75.8%). However, 11 participants were excluded due to poor-quality OCTA in both eyes and 1 was excluded due to bilateral carotid endarterectomies, leaving 261 participants included in the analysis (see Fig. 2). These 261 participants (mean ± SD age 72.0 ± 9.3 years, 57.1% male, median diabetes duration 15.4 years [IQR 11.1–22.4]) were not statistically different in terms of age (P = 0.11), sex (P = 0.29), or diabetes duration (P = 0.20) from the 99 who were invited but did not participate or were excluded. Compared with the remaining 1,290 FDS2 participants with type 2 diabetes who were not included in our analysis, the present sample of 261 was not statistically different in terms of sex (P = 0.07), but their age was an average of 3.6 years younger and they had a median 4.0 years shorter diabetes duration (P < 0.001) at FDS2 entry.

Both eyes in 213 participants and one eye in each of the remaining 48 participants were included, representing a total of 474 eyes. The 48 eyes excluded comprised 3 with ipsilateral endarterectomy and 45 with poor OCTA image quality. Any stenosis of the ipsilateral carotid artery was present in 58.0% (n = 275) of eyes, and 8.2% (n = 39) of these eyes had ipsilateral carotid stenosis ≥50%.

A higher parafoveal vessel density in the deep capillary plexus was associated with younger age, lower HbA_1c, shorter diabetes duration, lower ABSI, lower systolic blood pressure, and better renal function (see Table 1). Those with a higher parafoveal vessel density in the deep capillary plexus were less likely to be male, on insulin, have retinopathy, have any stenosis, and have IMT thickening in the CCA and at the bifurcation. In the unadjusted generalized estimating equations, all measures of carotid disease were significantly associated with a decreased parafoveal vessel density in the deep capillary plexus (Supplementary Table 2). In the most parsimonious model, age, sex, diabetes duration, and insulin use were significantly associated with a lower parafoveal vessel density in the deep capillary plexus (see Table 2). When the carotid disease variables were added one at a time to this model, ipsilateral ln(CCA) (coefficient −2.55 [95% CI −4.76, −0.35], P = 0.023) and the presence of any ipsilateral stenosis (−0.82 [−1.48, −0.17], P = 0.014) were statistically significantly associated with a lower parafoveal vessel density in the deep capillary plexus. The parafoveal vessel density in the superficial capillary plexus was significantly associated with some measures of carotid disease in the unadjusted models, but no ipsilateral carotid disease variables were significant when added to the most parsimonious model.

Associations seen in the bivariable analysis for FD-300 were similar to those for parafoveal vessel density in the deep capillary plexus (Supplementary Table 3). In the unadjusted generalized estimating equations, ipsilateral bifurcation IMT and ipsilateral stenosis were significantly associated with a lower FD-300. However, after adjustment for age, sex, and diabetes duration, only ipsilateral bifurcation IMT ≥1 mm (coefficient −0.79 [95% CI −1.50, −0.08], P = 0.030) was significantly and independently associated with FD-300 (Table 2). In the bivariable analysis, no measure of carotid disease was significantly associated with the FAZ. There were also no carotid disease variables significantly associated with the FAZ when added to the most parsimonious model, although ipsilateral ln(bifurcation IMT) (coefficient −0.025 [95% CI −0.047, −0.003], P = 0.028) was significantly associated with a lower FAZ area in the unadjusted model.

Some ipsilateral carotid disease variables were associated with flow area in the choriocapillaries and with retinal thickness in the unadjusted models (Supplementary Table 2). However, none were statistically significant when added to the most parsimonious models.

The left and right mean bifurcation IMT and CCA IMT were moderately correlated (r_s = 0.46, P < 0.001, and r = 0.46, P < 0.001, respectively). There was no statistically significant difference between the presence of any stenosis and the presence of bifurcation IMT ≥1 mm between the left and right sides (P = 0.064 and 0.731, respectively). However there was a significant difference between the presence of CCA ≥1 mm on the left and right sides (P = 0.001). The left and right OCTA variables were also all moderately correlated (r = 0.35–0.76, P < 0.001).

The significant associations between contralateral carotid disease–related variables and OCTA parameters are shown in Supplementary Table 2. When added to the most parsimonious models, contralateral ln(CCA) was statistically significantly associated with lower parafoveal vessel density in the deep capillary plexus (coefficient −1.97 [95% CI −3.67, −0.27], P = 0.023), contralateral stenosis was significantly associated with lower parafoveal vessel density in the superficial capillary plexus (coefficient −1.17 [95% CI −1.81, −0.54], P < 0.001), and contralateral CCA ≥1 mm was associated with lower FD-300 (coefficient −1.27 [95% CI −2.49, −0.04], P = 0.043).

This is the first study to show that carotid arterial disease is independently associated with lower retinal vessel density assessed by OCTA in people with type 2 diabetes, providing novel evidence of an association between macrovascular and microvascular complications. Ipsilateral and contralateral stenosis and wall thickening of the CCA and ipsilateral carotid wall thickening at the bifurcation were all independently associated with decreased density of microvessels surrounding the fovea. While OCTA is not routinely used to diagnose DR, a range of quantitative and qualitative measurements of OCTA are associated with the severity of retinopathy (10). Many studies have assessed the relationship between carotid disease and DR and, although there are those that did not find a statistically significant relationship (3,5,24,25), most reported a significant association, often between any DR and the presence of any stenosis (2,4,22,26–33). Our results provide support for these findings through access to a more sensitive retinal imaging modality than the color fundus photography conventionally used to identify DR.

Carotid procedures for high-grade stenosis have been shown to improve bilateral retinal and choroidal vessel density. A study involving 20 participants with severe stenosis found a significantly increased vessel density in the deep capillary plexus and improvement in visual field testing in both the ipsilateral and contralateral eye 4–5 weeks after carotid angioplasty and stenting (34). A pilot study showed significant improvement in the flow density in the radial peripapillary capillary network of the optic nerve head in both contralateral and ipsilateral eyes in 25 participants who had a carotid endarterectomy but found no statistically significant association with retinal vessel density in the deep or superficial capillary plexuses, perhaps due to lack of statistical power (35). Carotid endarterectomy has also been shown to improve bilateral subfoveal choroidal thickness (36). The present findings are consistent with previous reports in that we also detected a significant association between the ipsilateral and contralateral carotid disease variables and OCTA measures. There appears to be a consistent association between carotid disease and bilateral retinal vessel density.

The mechanism of how carotid disease affects retinal microvascular disease is not well established, although there are several hypotheses. Similar risk factors may contribute to both complications (1,33). Hyperglycemia, dyslipidemia, and hypertension are known risk factors for both atherosclerosis and retinopathy (1). As studies have shown improvement in retinal vasculature after surgical procedures for carotid disease (34–36), there may be a more direct association. The carotid artery supplies the ophthalmic and retinal arteries, and so carotid disease may attenuate blood flow to the eye. A previous study has shown that hemodynamic changes in the ophthalmic artery assessed by Doppler are associated with the presence of carotid plaque (37). Nevertheless, we found that both ipsilateral and contralateral carotid disease variables were associated with OCTA measures. While this may appear to support shared risk factors, it could also suggest that blood flow to the eye may be influenced by macrovascular disease in parts of the Circle of Willis other than the ipsilateral carotid artery. Contralateral cerebral hemodynamic improvement has been reported after carotid endarterectomy after a 10-year follow-up period, suggesting improved collateral circulation (38). Moreover, as previously mentioned, bilateral improvements in retinal microcirculation have been reported after carotid procedures with shorter follow-up (34–36).

The strengths of our study include comprehensive assessments in a relatively large community-based sample, use of novel technology to assess retinal vasculature, and statistical methodology that allowed the inclusion of both eyes with adjustment for their correlation and thus reduction of bias. Limitations include the inability to establish a causal relationship due to the cross-sectional nature of the study and implications for generalizability due to recruitment of a subgroup of surviving FDS2 participants. Although not the dependent variable in the current study, DR was ascertained from single-field fundus photography, and some cases of DR may have been missed or DR severity may have been misclassified. In addition, while we were able to correct the FAZ area, we were unable to correct the vessel density measurements for changes in image size magnification due to AL variation (39). We assessed the impact of AL on these OCTA measurements, but this was not an independent associate in the generalized estimating equations. Finally, flow signal in OCTA imaging is generated by rapidly moving blood cells within the capillaries. Reduction in flow within capillaries may simulate capillary dropout as demonstrated by “reversibility” of capillary loss in OCTA imaging. Our observation of reduced vessel density may reflect reduction in blood flow velocity due to increased blood viscosity, which has been reported in diabetes (40). Follow-up imaging is required for determination of whether “vessel density” can be improved with increased retinal perfusion velocity.

In conclusion, we found that ipsilateral and contralateral markers of carotid arterial disease in people with type 2 diabetes are associated with decreased retinal microvascular parameters assessed using OCTA. The mechanisms underlying this association warrant further study, especially given the clinical implications of using OCTA parameters as potential biomarkers for cardiovascular and cerebrovascular disease. Carotid disease is a useful clinical indicator of retinal microvascular disease as well as a marker of cardiovascular risk. The present data suggest that people with type 2 diabetes and carotid disease should be strongly encouraged to maintain regular ophthalmic screening and that appropriate intensive management of carotid disease may improve retinal microcirculation.

Acknowledgments. The authors thank study participants and the staff who assisted with data collection, specifically, Valentina Hellbusch, Penelope Dwyer, and Michelle England (Medical School, The University of Western Australia, Fremantle Hospital, Fremantle, Western Australia). The authors also thank both the Eye Clinic at Fremantle Hospital and Health Service and Lions Outback Vision for access to ophthalmological facilities for the duration of the study.

Funding. This project was funded by the Edith Hearn Bequest Grant awarded by the Spinnaker Health Research Foundation; SKG Radiology donated an ultrasound machine; and the Lions Eye Institute donated funds which contributed to purchase of the OCTA device. J.J.D. is supported by the Warren Jones/UWA Postgraduate Research Scholarship and Australian Government Research Training Program Scholarship. F.K.C. is supported by the Australian National Health and Medical Research Council Career Development Fellowship (MRF1142962). T.M.E.D. is supported by a Medical Research Future Fund Practitioner Fellowship.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.J.D. collected and analyzed the data and wrote the first draft of the manuscript. F.K.C. graded the retinal images, reviewed the ophthalmic data, provided clinical interpretation of the data, and reviewed and edited the manuscript. A.M.B. and B.T.D. reviewed the carotid ultrasound images, provided clinical interpretation of the data, and reviewed and edited the manuscript. A.W.T. provided clinical interpretation of the data and reviewed and edited the manuscript. T.M.E.D., the Principal Investigator of the FDS2, conceived the study, provided clinical interpretation, and produced the final version of the manuscript. W.A.D., a Co-Investigator of the FDS2, provided statistical advice and edited the manuscript. T.M.E.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the Australasian Diabetes Congress Annual Scientific Meeting, Sydney, Australia, 21–23 August 2019.