TYPE Original Research PUBLISHED 03 November 2022 DOI 10.3389/fcvm.2022.945421 Diabetic retinopathy predicts OPEN ACCESS cardiovascular disease EDITED BY Masanori Aikawa, Brigham and Women’s Hospital independently of subclinical and Harvard Medical School, United States atherosclerosis in individuals REVIEWED BY Shun Ishibashi, Jichi Medical University, Japan with type 2 diabetes: A Rafael Simó, Vall d’Hebron University Hospital, Spain Masahide Hamaguchi, prospective cohort study Kyoto Prefectural University of Medicine, Japan Esmeralda Castelblanco1,2† , Minerva Granado-Casas3,4† , *CORRESPONDENCE Emilio Ortega Marta Hernández5,6 , Montserrat Pinyol7 , Eudald Correig8 , eortega1@clinic.cat Josep Julve3,4 , Marina Idalia Rojo-López4 , Núria Alonso3,9 , Didac Mauricio didacmauricio@gmail.com Angelo Avogaro10 , Emilio Ortega11,12,13* and † These authors have contributed Didac Mauricio3,4,14* equally to this work and share first 1 authorship Division of Endocrinology, Metabolism and Lipid Research, Washington University School of Medicine in St. Louis, St. Louis, MO, United States, 2 DAP-Cat Group, Unitat de Suport a la Recerca SPECIALTY SECTION Barcelona, Fundació Institut Universitari per a la Recerca a l’Atenció Primària de Salut Jordi Gol i This article was submitted to Gurina, Barcelona, Spain, 3 Center for Biomedical Research on Diabetes and Associated Metabolic Atherosclerosis and Vascular Medicine, Diseases (CIBERDEM), Instituto de Salud Carlos III, Barcelona, Spain, 4 Department of Endocrinology a section of the journal and Nutrition, Hospital de la Santa Creu i Sant Pau and Sant Pau Biomedical Research Institute (IIB Frontiers in Cardiovascular Medicine Sant Pau), Barcelona, Spain, 5 Lleida Institute for Biomedical Research Dr. Pifarré Foundation IRBLleida, University of Lleida, Lleida, Spain, 6 Department of Endocrinology and Nutrition, University RECEIVED 16May 2022 Hospital Arnau de Vilanova, Lleida, Spain, 7 Consorcio de Atención Primaria del Eixample (CAPSE), ACCEPTED 19 October 2022 Grup Transversal de Recerca en Atenció Primària, Institut d’Investigacions Biomédiques August Pi i PUBLISHED 03 November 2022 Sunyer (IDIBAPS), Barcelona, Spain, 8 Department of Biostatistics, Universitat Rovira i Virgili, Reus, CITATION Spain, 9 Department of Endocrinology and Nutrition, Germans Trias i Pujol Hospital and Research Castelblanco E, Granado-Casas M, Institute, Universitat Autònoma de Barcelona, Badalona, Spain, 10 Department of Medicine, Hernández M, Pinyol M, Correig E, Università di Padova, Padua, Italy, 11 Diabetes Unit, Department of Endocrinology and Nutrition, Julve J, Rojo-López MI, Alonso N, Hospital Clínic de Barcelona, Barcelona, Spain, 12 Center for Biomedical Research on Avogaro A, Ortega E and Mauricio D Pathophysiology of Obesity and Nutrition (CIBEROBN), Instituto de Salud Carlos III (ISCIII), (2022) Diabetic retinopathy predicts Barcelona, Spain, 13 Institut d’Investigacions Biomèdiques August Pi i Sunyer, Barcelona, Spain, 14 cardiovascular disease independently Faculty of Medicine, University of Vic (UVic-UCC), Barcelona, Spain of subclinical atherosclerosis in individuals with type 2 diabetes: A prospective cohort study. Front. Cardiovasc. Med. 9:945421. Background: Diabetic retinopathy (DR) and preclinical atherosclerosis are doi: 10.3389/fcvm.2022.945421 associated with higher cardiovascular risk. However, no studies have COPYRIGHT investigated the predictive role of DR and preclinical atherosclerosis jointly © 2022 Castelblanco, Granado-Casas, Hernández, Pinyol, Correig, Julve, on cardiovascular events in subjects with type 2 diabetes (T2D). We aimed Rojo-López, Alonso, Avogaro, Ortega to assess the contribution of DR and subclinical atherosclerosis on the and Mauricio. This is an open-access article distributed under the terms of risk of adverse cardiovascular events in subjects with T2D without previous the Creative Commons Attribution cardiovascular disease (CVD). License (CC BY). The use, distribution or reproduction in other forums is Methods: We included two prospective cohorts of subjects with T2D from permitted, provided the original author(s) and the copyright owner(s) the same geographical area. Assessment of subclinical atherosclerosis was are credited and that the original performed by carotid ultrasound. An ophthalmologist classified DR according publication in this journal is cited, in to standard criteria. Cardiovascular outcomes considered for analysis were accordance with accepted academic practice. No use, distribution or the following: ischemic heart disease, stroke, heart failure, peripheral artery reproduction is permitted which does disease, revascularization procedures, and cardiovascular mortality. Bivariable not comply with these terms. and multivariable predictive models were performed. Frontiers in Cardiovascular Medicine 01 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 Results: From a total of 374 subjects with T2D 44 developed cardiovascular events during the 7.1 years of follow-up. Diabetes duration, total cholesterol, and glycated hemoglobin (HbA1c) at baseline were higher in subjects who developed cardiovascular outcomes (p < 0.001, p = 0.026, and p = 0.040, respectively). Compared with subjects without events, those developing cardiovascular events had higher prevalence of retinopathy (65.9% vs. 38.8%, p = 0.001; respectively) and more than mild retinopathy (43.2% vs. 31.8%, p = 0.002; respectively). Furthermore, all-cause mortality was higher in subjects with MACE than those without events (13.6% vs. 3.3%, p = 0.009; respectively). The multivariable analyses showed that HbA1c and the presence of DR at baseline were predictive of cardiovascular outcomes (p = 0.045 and p = 0.023, respectively). However, the burden of subclinical atherosclerosis was not (p = 0.783 and p = 0.071, respectively). Conclusion: DR is a strong predictor of cardiovascular events in T2D individuals at primary CVD prevention, even after accounting for the presence of preclinical carotid atherosclerosis. These results may help to individualize CVD prevention strategies in T2D. KEYWORDS type 2 diabetes, diabetic retinopathy, subclinical atherosclerosis, cardiovascular disease, major adverse cardiovascular events Introduction infarction and stroke events were associated with the presence and severity of DR in this population (12). The World Health Type 2 diabetes mellitus is associated with a higher risk of Organization (WHO) multinational study group reported that cardiovascular disease (CVD) (1). The most common clinical DR is positively associated with cardiovascular mortality and expression of CVD is coronary heart disease, cerebrovascular incident myocardial infarction in individuals with T2D (22). disease, peripheral artery disease, and congestive heart failure Moreover, other studies found that subjects with T2D with (2). These are often manifested as major adverse cardiovascular a mild or moderate DR had a higher risk of coronary heart events with hospitalizations, procedures, and deaths derived disease, stroke, and any CVD (17, 18, 26, 27). A meta-analysis from acute coronary syndromes, myocardial infarction, and of observational studies found a higher risk of heart failure and stroke, as well as sudden death (1, 2). According to the American stroke in individuals with DR (16). In addition, a prospective Diabetes Association (ADA), diabetic retinopathy (DR) is the study performed with individuals with T2D found that DR most common microvascular and neurovascular complication was a predictor of cardiovascular events independently of (3). DR is an independent predictor of subclinical CVD in type classical risk factors (21). Moreover, a recent retrospective 2 diabetes (T2D) (4). Furthermore, DR is associated with CVD cohort study performed with Catalan population found that in patients with T2D (5). Similarly, subclinical atherosclerosis DR was associated with coronary heart disease, cardiovascular is an independent predictor of cardiovascular events and death events, and all-cause mortality among subjects with T2D (20). in subjects with type 2 diabetes (6, 7). Besides, DR alone or Furthermore, a recent meta-analysis reported that DR in combination with other microvascular complications has increases the risk of cardiovascular death in patients with been independently associated with the presence of carotid diabetes (30). However, a post hoc analysis from a randomized atherosclerosis in this population (8–10). clinical trial did not show any association between DR and A significant number of studies were performed to assess increased risk of recurrent cardiovascular events in subjects the relative risk of CVD with the presence of DR in subjects with T2D (14). Nevertheless, other post hoc analyses from two with diabetes (11–29). Retrospective studies conducted with cardiovascular outcome clinical trials found an increased risk of large T2D samples found that the CVD risk increases with the CVD associated with microvascular complications (15). cumulative burden of microvascular complications, i.e., DR, On the other hand, subclinical atherosclerosis has been diabetic nephropathy, and neuropathy (11, 13). Furthermore, a associated with a higher risk of CVD in subjects with T2D (6, recent prospective study determined that incident myocardial 7, 31, 32). Specifically, carotid plaque calcification predicted Frontiers in Cardiovascular Medicine 02 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 cardiovascular events in T2D subjects with a low degree of informed consent was obtained from each patient included in stenosis (6). Furthermore, carotid plaques were predictors of the study. The study protocol conforms to the ethical guidelines cardiovascular morbidity in individuals with T2D without of the 1975 Declaration of Helsinki, and the study has been prior CVD (7). Moreover, the presence of carotid plaques was priorly approved by both Institution’s ethics committees on more predictive for prevalent silent coronary atherosclerosis, research on humans. severity, and extent in asymptomatic T2D individuals (31); besides, intima-media thickness (IMT) and carotid plaques were independent predictors of CVD (32). However, the role Clinical variables of subclinical atherosclerosis in CVD risk assessment is still unclear (33). A description of the procedures to collect clinical variables of To our knowledge, there are no reports investigating the two cohorts has been previously published (10, 34). Age, sex, together in the same study the potential predictor role of DR and tobacco exposure (current or former smoking), systolic blood subclinical atherosclerosis on the development of cardiovascular pressure (sBP), diastolic blood pressure (dBP), and diabetes events in subjects with T2D. We hypothesized that as the duration were collected. Hypertension and dyslipidemia were presence of subclinical atherosclerosis involves detecting the defined as active treatment with antihypertensive and lipid- etiological process leading to atherosclerotic cardiovascular lowering agents, respectively. Weight, height, waist, and blood events, DR may contribute to further predicting CVD in type pressure were measured by standard procedures. Blood and 2 diabetes. Therefore, we aimed to assess the predictive role urine samples were collected in the fasting state to obtain of subclinical atherosclerosis and DR on the development of biochemical parameters using standardized methods. The cardiovascular outcomes in subjects with T2D without prior estimated glomerular filtration rate (eGFR) was calculated CVD without chronic kidney disease. using The Modification of Diet in Renal Disease (MDRD- 4) formula (35). A thorough revision of the medical records, including all the available information from the health Materials and methods system was performed to identify death and its causes and cardiovascular events during the follow-up period; these Two cohorts of individuals with T2D from the same included ischemic heart disease (including any recorded geographical area recruited in two previous studies were diagnosis of any form of ischemic heart disease, and angina included (10, 34). These subjects were prospectively followed pectoris), stroke, heart failure, peripheral artery disease, to determine incident cardiovascular events. The first cohort procedures of revascularization, and cardiovascular mortality. (Cohort 1) was previously selected to study the relationship The cardiovascular events were only confirmed when the between carotid atherosclerosis and diabetic retinopathy in treating physician had recorded a new diagnosis corresponding T2D. From 312 subjects recruited in this cohort, we included 310 to that given event after hospital admission or, alternatively, in subjects with complete eye and carotid ultrasound assessments the outpatient medical records. from this cohort. The characteristics of the individuals were detailed in a previous publication (10). Briefly, the inclusion criteria in this study were as follows: diagnosis of T2D; age Diabetic retinopathy from 40 to 75 years; absence of a prior history of CVD; absence of impaired renal function defined as a estimated glomerular The study participants underwent a complete baseline filtration rate (eGFR) < 60 ml/min or macroalbuminuria; no examination by an ophthalmologist. According to the previous history of diabetic foot disease; absence of major international consensus on DR (36), DR was assessed with psychiatric disorders. The second cohort (Cohort 2) was a multifield stereoscopic retinal photography. Moreover, this recruited to investigate the frequency of preclinical carotid was classified as follows: (a) no DR; (b) mild non-proliferative atherosclerosis (the DIABIMCAP Study) (34). From a cohort DR if only microaneurysms were observed; (c) moderate of 106 subjects with T2D included in the DIABIMCAP Study, non-proliferative DR if more than just microaneurysms but a total sample of 64 subjects with eye and carotid ultrasound less than severe non-proliferative DR were detected; (d) severe data were included in the current study. A detailed description non-proliferative DR if any of the following characteristics of this cohort can be found elsewhere (34). The inclusion were observed: over 20 intraretinal hemorrhages in each of 4 criteria were a diagnosis of T2D within the previous year; the quadrants, definite venous beading in 2 + quadrants, prominent absence of prior CVD, cancer, chronic renal or liver disease; intraretinal microvascular abnormalities in 1 + quadrant, as history of alcohol or drug abuse; major psychiatric or chronic well as no signs of proliferative DR; and (e) proliferative DR if illness, and short life expectancy. Therefore, a final sample of neovascularization and/or vitreous/preretinal hemorrhage was 374 individuals with T2D was prospectively followed. Written detected (10, 34). Frontiers in Cardiovascular Medicine 03 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 TABLE 1 Clinical characteristics of the study group according to TABLE 2 Regularized logistic regression models for the association incident cardiovascular events. between the different variables, including plaque and diabetic retinopathy, and incident major adverse cardiovascular events. Characteristics No CVE CVE p N = 328 N = 46 Variables OR (95% CI) p OR (95% CI) P Sex (women) 150 (45.7) 26 (56.5) 0.224 Intercept 0.09 (0.07–0.12) <0.001 0.10 (0.08–0.12) <0.001 Sex (women) 1.11 (0.93–1.34) 0.107 1.11 (0.94–1.31) 0.138 Age (years) 60.5 [52.0–67.0] 60.0 [54.0–69.0] 0.366 Age (years) 1.03 (0.94–1.12) 0.470 1.03 (0.95–1.11) 0.449 Hypertension 185 (56.4) 31 (67.4) 0.210 Hypertension 1.12 (0.96–1.34) 0.265 1.14 (0.98–1.32) 0.254 Dyslipidemia 142 (43.3) 27 (58.7) 0.071 Dyslipidemia 1.12 (0.97–1.30) 0.390 1.13 (0.99–1.29) 0.339 Waist (cm) 104.0 106.0 0.116 Diabetes 1.08 (0.98–1.19) 0.439 1.08 (0.99–1.18) 0.323 [97.0–111.0] [99.8–114.0] duration (years) BMI (kg/m2 ) 30.3 [27.7–34.2] 32.0 [28.8–35.3] 0.113 Waist (cm) 1.06 (0.98–1.14) 0.506 1.06 (0.99–1.13) 0.539 Tobacco exposure 184 (56.1) 26 (56.5) 0.511 Tobacco 1.05 (0.86–1.27) 0.124 1.05 (0.88–1.25) 0.132 Diabetes duration (years) 5.4 [2.0–10.0] 11.0 [5.3–20.0] <0.001 exposure sBP (mmHg) 137.0 142.0 0.051 HbA1c (%) 1.09 (1.00–1.18) 0.045 1.09 (1.02–1.17) 0.034 [126.0–148.0] [130.0–156.0] Albumin/ 1.06 (0.98–1.14) 0.870 1.06 (0.99–1.13) 0.815 creatinine ratio dBP (mmHg) 78.0 [71.5–86.0] 78.0 [69.0–85.8] 0.471 (mg/g) Total cholesterol (mg/dl) 183.0 193.0 0.030 eGFR 1.03 (0.95–1.11) 0.386 1.03 (0.96–1.10) 0.338 [163.0–212.0] [174.0–223.0] (mL/min/1.73 HDL-cholesterol (mg/dl) 48.0 [40.8–58.0] 48.0 [45.0–63.0] 0.086 m2 ) LDL-cholesterol (mg/dl) 108.0 112.0 0.345 Subclinical 1.06 (0.90–1.25) 0.833 1.06 (0.92–1.23) 0.783 [88.4–130.0] [95.8–133.0] carotid atherosclerosis Triglycerides (mg/dl) 118.0 128.0 0.635 [87.0–167.0] [81.2–171.0] Diabetic 1.27 (1.06–1.52) 0.023 - - retinopathy HbA1c (%) 7.3 [6.6–8.3] 7.8 [6.9–9.4] 0.023 Diabetic - - 1.27 (0.92–1.76) 0.074 Albumin/creatinine ratio 7.0 [3.9–14.7] 12.0 [5.0–25.0] 0.027 retinopathy (mg/g) (severe) eGFR (mL/min/1.73 m2 ) 90.4 91.1 0.376 AUC = 0.69 for the model including the presence of subclinical atherosclerosis [77.7–104.0] [85.4–104.0] and retinopathy. AUC = 0.68 for the model including the presence of subclinical Subclinical carotid 203 (61.9) 32 (69.6) 0.538 atherosclerosis and severity of retinopathy. Severe diabetic retinopathy group compared atherosclerosis with the group with mild or no DR. CI, confidence interval; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin; OR, odds ratio. The bold values are statistical Categorized subclinical 0.188 significants with p values lower than 0.05. carotid aterosclerosis No 125 (38.1) 14 (30.4) One plaque 86 (26.2) 8 (17.4) Carotid ultrasound imaging Multiple plaques 117 (35.7) 24 (52.2) Advanced aterosclerosis 60 (18.2) 15 (34.1) 0.014 All the study subjects underwent the same carotid Diabetic retinopathy 126 (38.4) 31 (67.4) 0.001 ultrasound imaging protocol. Bilateral carotid artery B-mode DR status 0.001 ultrasound imaging to assess IMT and the presence of No 202 (61.6) 15 (32.6) atherosclerotic plaques was performed following the Mild 56 (17.1) 10 (21.7) standardized protocol (10, 34). A semiautomatic software Moderate or severe 70 (21.3) 21 (45.7) measured IMTs of the common carotid artery (CCA), bulb, and Deaths 9 (2.7) 8 (17.4) <0.001 internal carotid artery (ICA). IMT-mean and IMT-maximum Cardiovascular mortality 0 (0.0) 3 (6.5) 0.002 from each carotid segment were collected. Atherosclerotic Ischemic heart disease 0 (0.0) 24 (52.2) <0.001 plaques were assessed by using B-mode and color Doppler; Stroke 0 (0.0) 9 (19.6) <0.001 these were defined as a focal wall thickening encroaching into Heart failure 0 (0.0) 13 (28.3) <0.001 the arterial lumen by at least 50% of the surrounding IMT value Peripheral artery disease 0 (0.0) 11 (23.9) <0.001 or with a thickness of 1.5 mm at minimum measuring from Procedure of 0 (0.0) 12 (26.1) <0.001 the media adventitia interference to the intima-lumen surface, revascularization following the Mannheim consensus (37). Data are shown as n (%) for categorical variables and median [interquartile range] for continuous variables. Multiple plaques include two or more plaques. Advanced atherosclerosis consists of the presence of three or more plaques. CVE, cardiovascular Statistical methods events; DR, diabetic retinopathy; dBP, diastolic blood pressure; BMI, body mass index; eGFR, estimated glomerular filtration rate; HbA1c, glycated hemoglobin; HDL-cholesterol, high-density lipoprotein-cholesterol; LDL-cholesterol, low-density Continuous variables were tested for normality using the lipoprotein-cholesterol; sBP, systolic blood pressure. Shapiro–Wilk test. Data are presented as median and 25th Frontiers in Cardiovascular Medicine 04 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 and 75th percentiles for continuous variables with a non- a graphical representation of the conditional independence normal distribution or the mean and standard deviation (SDs) relationships among all the variables, allowing us to infer causal for variables with a normal distribution. Categorical variables relationships between predictor and predicted variables. are reported as percentages. Differences between groups were All statistical analyses were performed using the R software analyzed using the non-parametric Mann–Whitney U-test, package version 4.0 (40). Welch’s parametric t-test for continuous variables, and the chi-square test or Fisher’s exact test for categorical variables. All continuous variables were standardized and normalized Results whether necessary. We included subjects with T2D from two different cohorts Clinical characteristics of the study subjects are shown in in the current study. As these patients harbor a different baseline Table 1. After a median (SD) follow-up of 7.1 (1.0) years, risk, we could not perform Kaplan-Meier and Cox regressions, among the 374 participants, 46 reported cardiovascular events. as they are not appropriate to estimate the survival function in This represents a 1.2% annual rate of cardiovascular events this setting. Therefore, we carried out a series of multivariate in this primary prevention cohort. Subjects who reported predictive models to assess the relationship between prevalent cardiovascular outcomes had a longer diabetes duration plaque and retinopathy and the cardiovascular outcome. As for (p < 0.001), higher total cholesterol (p = 0.030), and glycated the exposure, we divided plaque into Yes (1 or more plaques) hemoglobin (HbA1c) (p = 0.023) in comparison with those and No, or as Multiple plaques (two or more) or Advanced who remained free of cardiovascular events at the end of the atherosclerosis (three or more plaques). Moreover, DR was follow-up. Furthermore, they had also a higher prevalence of DR classified into either Yes or No, or into No or Mild vs. Severe. For (67.4% vs. 38.4%, p = 0.001; respectively) and DR severity (45.7% each set of models, we took a dual approach to assess the effect vs. 21.3%, p < 0.001; respectively). In addition, mortality was of the predictive variables on the target variable. Random forests higher in the group with cardiovascular events (17.4% vs. 2.7%, prove to be more accurate models for such complex scenarios p < 0.001; respectively). The description of the study sample (38) and, as an added advantage, allowed us to assess the relative according to the presence of atherosclerotic plaques and DR importance of each variable via out-of-bag accuracy. However, is shown in Supplementary Table 1. Those participants with the downside is that random forests do not allow to evaluate plaque(s) were older (p < 0.001), had higher sBP (p = 0.010) whether each variable has a protective or harmful effect. For this and higher frequency of hypertension (p = 0.034), tobacco reason, we used the result of the regularized regression through exposure (p < 0.001) compared with individuals without which we can evaluate the direction of the effect through the atherosclerosis. On the other hand, individuals with DR had a negativity or positivity of the coefficient. Furthermore, in the higher frequency of hypertension (p = 0.007), atherosclerotic logistic regression models, we imposed a highly constrained plaques (p = 0.014), deaths (p = 0.028), cardiovascular mortality regularization that forces model coefficients to be as small as (p = 0.030), and cardiovascular events (p = 0.001) in comparison possible, sometimes even 0. Therefore, if two variables were with the non-DR group. Moreover, they had higher waist closely associated with each other, the model removes the effect (p = 0.003), diabetes duration (p < 0.001), sBP (p < 0.001), of the least important one. In the random forests, we constrained HDL-cholesterol (p = 0.021), and albumin/creatinine ratio the ability of the trees to generate branches, thus reducing the (p < 0.001). The clinical characteristics of both study cohorts effect of colinear variables. In both cases, the models identified are shown in Supplementary Table 2. Participants from Cohort the variable of interest as the most important one in explaining 1 were older (p = 0.030) and had a lower frequency of the outcome. Furthermore, in the logistic regression models, tobacco exposure (p < 0.001) in comparison with those from we imposed a highly constrained regularization that forces Cohort 2. Furthermore, these participants had a higher diabetes model coefficients to be as small as possible, sometimes even duration (p < 0.001), sBP (p = 0.032), HbA1c (p < 0.001), 0. Therefore, if two variables were closely associated with each eGFR (p < 0.001), and a higher frequency and severity of DR other, the model removes the effect of the least important one. (p < 0.001). However, they showed lower dBP (p < 0.001), In the random forests, we constrained the ability of the trees to total cholesterol (p = 0.007), LDL-cholesterol (p = 0.003), generate branches, thus reducing the effect of colinear variables. and presence of subclinical carotid atherosclerosis (p < 0.001) In both cases, the models identified the variable of interest compared with Cohort 2. as the most contributing one in explaining the outcome. We provide odds ratios for the logistic regressions and estimates for the linear regressions, both with 95% confidence intervals and Diabetic retinopathy, preclinical p-values. atherosclerosis, and cardiovascular Finally, to study further the etiology of our study, we events also built a Bayesian Network among the variables using the PC algorithm (39). This algorithm is able to compute The multivariate predictive model for cardiovascular events causal relationships between variables. This network provides showed that both HbA1c (1% change) and the presence of Frontiers in Cardiovascular Medicine 05 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 FIGURE 1 Random forest of the relative importance of each variable in the multivariate model for the association between the presence of atherosclerotic plaque and diabetic retinopathy with cardiovascular events. HbA1c, glycated hemoglobin; DM_duration, diabetes mellitus duration; Plaque, presence of carotid plaques; eGFR, estimated glomerular filtration rate. DR were predictors of cardiovascular events [OR 1.09, 95% events [OR 1.41 95% CI (1.12–1.77); p = 0.091, and OR 1.44 CI (1.00–1.18), p = 0.045 and OR 1.27, 95% CI (1.06–1.52), (1.16–1.78); p = 0.071, respectively] in multivariate models p = 0.023; respectively] after accounting for confounders (Supplementary Table 3). (Table 2). However, the presence of subclinical carotid atherosclerotic plaques was not associated with cardiovascular events [OR 1.06, 95% CI (0.90–1.25), p = 0.833]. In the relative Causal inference on cardiovascular variable importance of this model, HbA1c and DR had a higher events mean decrease accuracy being the two most important variables of this model (Figure 1). The prediction causation algorithm to explore the causal On the other hand, when considering in this model a more relationship between DR, subclinical carotid atherosclerosis advanced definition of DR, preclinical atherosclerosis did not and cardiovascular outcomes showed that severe DR had a have a predictive role in the development of cardiovascular causal solid connection for the development of cardiovascular events either [OR 1.06, 95% CI (0.92–1.23); p = 0.783] (Table 2). events (Figure 3). Subclinical atherosclerosis was not related to In this case, severe DR was not significantly associated with cardiovascular events in our study sample. Nevertheless, female cardiovascular events [OR 1.27, 95% CI (0.92–1.76); p = 0.074]. sex, obesity, and tobacco exposure were associated with the Besides, diabetes duration and HbA1c had a decisive role development of severe atherosclerosis, defined as the presence of in this multivariate prediction model, as shown in Figure 2. three or more plaques. In addition, all-cause mortality was not Nevertheless, HbA1c had a risk association of cardiovascular associated with any of the other clinical variables even though, events [OR 1.09, 95% CI (1.02–1.17), p = 0.034]. Furthermore, as expected, classical cardiovascular risk factors were related to the presence of DR or advanced atherosclerosis (three or the increasing albumin/creatinine ratio and, consequently, the greater plaques) were not predictive factors of cardiovascular presence of severe DR. Frontiers in Cardiovascular Medicine 06 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 FIGURE 2 Random forest of the relative importance of each variable in the multivariate predictive model for the association between severe diabetic retinopathy and the presence of subclinical atherosclerosis with cardiovascular events. DM_duration, diabetes mellitus duration; eGFR, estimated glomerular filtration rate; Plaque, presence of carotid plaques. Diabetic retinopathy severe, reference group: no or mild diabetic retinopathy. Discussion a post hoc analysis from a clinical trial did not find any association between DR or neuropathy with cardiovascular This is the first study that assessed the independent death, non-fatal myocardial infarction, stroke, hospitalization predictive role of DR and subclinical atherosclerosis in for unstable angina, and recent acute coronary syndrome (14). developing cardiovascular events in T2D individuals without a Nevertheless, the authors identified the diabetes duration as history of the previous CVD at baseline and established chronic an independent predictor of cardiovascular events beyond DR, kidney disease. Our findings showed that HbA1c and DR, but neuropathy, and cardiovascular risk factors, which is in line not preclinical atherosclerosis, were predictors of cardiovascular with our findings on the importance of the disease duration in events considering the presence of subclinical atherosclerosis. the predictive models. In addition, the WHO study researchers The presence of DR and a poorer glycemic control were found that diabetes duration and glycemic control predicted associated with a higher risk of developing a cardiovascular incident fatal and non-fatal myocardial infarction in subjects event in subjects with T2D, adjusting for classical risk factors with diabetes (22). Therefore, the role of DR, glycemic control, and considering the presence and burden of subclinical and diabetes duration in predicting cardiovascular events should atherosclerosis; this is in line with previous studies (20, 21). be considered in this population. The potential relationship The burden of microvascular complications such as DR, between DR and subclinical atherosclerosis has been published nephropathy, and neuropathy was associated with a cumulative in previous studies (8–10, 41); additionally, the American Heart risk of cardiovascular death, non-fatal myocardial infarction, Association (AHA) and the European Society of Cardiology or non-fatal ischemic stroke in subjects with T2D (11). (ESC) included DR as an independent risk factor to predict Furthermore, microvascular complications were independently cardiovascular events in patients treated in primary prevention related to the 10-year risk of death and cardiovascular events (42, 43). in individuals with T2D (13). A recent systematic review On the other hand, subclinical atherosclerosis did not reported that DR is a strong predictor of stroke and CVD, predict a higher risk of developing cardiovascular events in suggesting that patients with DR have an overall worse prognosis subjects with T2D; however, the high prevalence (63%) of than those without this condition (5). On the other hand, patients with subclinical atherosclerosis in our study sample Frontiers in Cardiovascular Medicine 07 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 FIGURE 3 The prediction causation algorithm for the causal relationship between diabetic retinopathy, subclinical atherosclerosis and cardiovascular events in subjects with type 2 diabetes. CVE, cardiovascular events; Multiple plaques, multiple carotid plaques (two or more); sBP, systolic blood pressure; eGFR, estimated glomerular filtration rate; DM_duration, diabetes mellitus duration; HbA1c, glycated hemoglobin. may be a limitation to assess its predictive role, as well because microangiopathy and carotid atherosclerosis are closely as that of other highly prevalent classical risk factors such linked (8). as age, dyslipidemia, and hypertension. The presence of One of the limitations of our study is that we used two atherosclerotic plaques predicts future cardiovascular events different cohorts to establish the predictive role of DR and and mortality in patients with T2D (6, 7), a finding that is in subclinical atherosclerosis in the development of cardiovascular contrast with our results. However, participants in previously events. However, regularized logistic regression models were published studies had other late diabetic complications, such performed to calculate the study participants’ risk. Although as nephropathy, that could strongly influence the outcome. the causal relationships between cardiovascular events and Individuals with a higher burden of plaques have a higher predictors were calculated using a Bayesian method, we cannot incidence of renal endpoints (7); moreover, carotid plaques are conclude this causal relationship due to the study design. predictors of renal outcomes in subjects with T2D. Nevertheless, Moreover, definitive conclusions cannot be established due microvascular complications and subclinical atherosclerosis to the rather insufficient number of incident cardiovascular may be closely linked in their respective pathogenetic pathways events and the short follow-up study period of the participants; (19). Our findings identify DR as a causal predicting factor in for this reason, the burden of subclinical atherosclerosis and developing a cardiovascular events in the prediction causation severe DR, as well as the other classical risk factors such algorithm; however, this cannot be definitively established as age, dyslipidemia, and hypertension did not have either a Frontiers in Cardiovascular Medicine 08 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 significant association with cardiovascular events. Moreover, conducted the study. ECo analyzed the data. EC, MG-C, AA, classical risk factors such as hypertension and dyslipidemia EO, and DM interpreted the results. EC and MG-C wrote the could be associated with each other which could influence manuscript. EO and DM were guarantors of this work and as the results. Furthermore, any association of cardiovascular such had full access to all the data in the study and take full events with HbA1c and DR would have not been found if the responsibility for the integrity of the data, and the accuracy traditional definition of major adverse cardiovascular events of the data analysis. All authors have revised the manuscript (MACE, i.e., non-fatal myocardial infarction, non-fatal stroke critically for important intellectual content and given final and cardiovascular death) were used as the total number of approval of the version to be published. events would have been even lower. In addition, data from study subjects on their dietary pattern, physical activity and alcohol consumption were not available. Therefore, their potential Funding contribution of these factors to the main outcome of the study could not be included in the models. On the other hand, this This study was supported by grants PS09/01035 and is the first prospective study that analyzed the predictive role of PI11/01723 from the Instituto de Salud Carlos III, Ministry DR and subclinical atherosclerosis on the cardiovascular events. of Economy and Competitiveness and FEDER, Spain. CIBER Therefore, these results may be highly relevant for the design for Diabetes and Associated Metabolic Diseases (CIBERDEM) of further studies focused on this issue. Furthermore, this study and CIBEROBN were initiatives of Instituto de Salud Carlos included a well-defined sample including participants without III, Spain. MR-L held a predoctoral fellowship from the other diabetic complications and no history of previous CVD, Departament de Salut, SLT017/20/000107. JJ was recipient allowing us specifically assessing the predictive role of DR and of a Miguel Servet Type 2 contract (CPII18/00004; ISCIII). subclinical atherosclerosis. JJ was member of Red de Investigación en “Enfermedades Metabólicas y Cáncer” (RED2018-102799-T), Ministerio de Economía y Competitividad (MINECO), Madrid, Conclusion Spain. Institut de Recerca del Hospital de la Santa Creu i Sant Pau and Biomedical Research Institute’s Dr. Pifarré In conclusion, the presence of DR was a strong predictor Foundation from Lleida (IRBLleida) were accredited by of CVD independently from the presence of subclinical carotid the Generalitat de Catalunya as Centre de Recerca de atherosclerosis in subjects with T2D without previous clinical CVD. Our findings point to a solid linkage between DR and Catalunya (CERCA). future CVD in T2D, which aligns with the published studies and guidelines in this issue. However, more extensive studies are needed to confirm the current findings. Conflict of interest The authors declare that the research was conducted in the Data availability statement absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation. Publisher’s note Ethics statement All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated The studies involving human participants were reviewed organizations, or those of the publisher, the editors and the and approved by the University Hospital Arnau de Vilanova, reviewers. Any product that may be evaluated in this article, or Lleida, Spain (CEIC 7/2011) and University Hospital Clinic claim that may be made by its manufacturer, is not guaranteed de Barcelona, Barcelona, Spain (CEIC 2011/6466). The or endorsed by the publisher. patients/participants provided their written informed consent to participate in this study. Supplementary material Author contributions The Supplementary Material for this article can be EC, MG-C, EO, and DM contributed to the conception and found online at: https://www.frontiersin.org/articles/10.3389/ design of the study. EC, MG-C, MH, MP, JJ, MR-L, and NA fcvm.2022.945421/full#supplementary-material Frontiers in Cardiovascular Medicine 09 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 References 1. Viigimaa M, Sachinidis A, Toumpourleka M, Koutsampasopoulos K, Alliksoo 19. Barlovic DP, Harjutsalo V, Groop PH. The association of severe diabetic S, Titma T. Macrovascular complications of type 2 diabetes mellitus. Curr Vasc retinopathy with cardiovascular outcomes in long-standing type 1 diabetes: a Pharmacol. (2019) 18:110–6. doi: 10.2174/1570161117666190405165151 longitudinal follow-up. Diabetes Care. (2019) 42:E49–50. doi: 10.2337/dci18-0058 2. International Diabetes Federation. IDF Atlas 9th Edition and Other 20. Barrot J, Real J, Vlacho B, Romero-Aroca P, Simó R, Mauricio D, et al. Diabetic Resources (2019). Available online at: https://diabetesatlas.org/en/resources/ retinopathy as a predictor of cardiovascular morbidity and mortality in subjects (accessed on Sep 21, 2021). with type 2 diabetes. Front Med. (2022) 9:945245. doi: 10.3389/fmed.2022.945245 3. Solomon S, Chew E, Duh E, Sobrin L, Sun J, VanderBeek B, et al. Diabetic 21. Planas A, Simó-Servat O, Hernández C, Ortiz-Zúñiga Á, Marsal JR, Herance retinopathy: a position statement by the American diabetes association. Diabetes JR, et al. Diabetic retinopathy and skin tissue advanced glycation end products are Care. (2017) 40:412–8. doi: 10.2337/DC16-2641 biomarkers of cardiovascular events in type 2 diabetic patients. J Pers Med. (2021) 11:1344. doi: 10.3390/JPM11121344 4. Simó R, Bañeras J, Hernández C, Rodríguez-Palomares J, Valente F, Gutierrez L, et al. Diabetic retinopathy as an independent predictor of subclinical 22. Fuller JH, Stevens LK, Wang S-L, Fuller J. Risk factors for cardiovascular cardiovascular disease: baseline results of the precised study. BMJ Open Diabetes mortality and morbidity: the WHO multinational study of vascular disease in Res Care. (2019) 7:e000845. doi: 10.1136/bmjdrc-2019-000845 diabetes. Diabetologia. (2001) 44(Suppl 2.):S54–64. 5. Pearce I, Simó R, Lövestam-Adrian M, Wong DT, Evans M. Association 23. Garofolo M, Gualdani E, Giannarelli R, Aragona M, Campi F, Lucchesi between diabetic eye disease and other complications of diabetes: implications for D, et al. Microvascular complications burden (nephropathy, retinopathy and care. a systematic review. Diabetes Obes Metab. (2019) 21:467–78. doi: 10.1111/ peripheral polyneuropathy) affects risk of major vascular events and all-cause dom.13550 mortality in type 1 diabetes: a 10-year follow-up study. Cardiovasc Diabetol. (2019) 18:159. doi: 10.1186/s12933-019-0961-7 6. De Kreutzenberg SV, Fadini GP, Guzzinati S, Mazzucato M, Volpi A, Coracina A, et al. Carotid plaque calcification predicts future cardiovascular events in type 2 24. Melo LGN, Morales PH, Drummond KRG, Santos DC, Pizarro MH, Barros diabetes. Diabetes Care. (2015) 38:1937–44. doi: 10.2337/dc15-0327 BSV, et al. Diabetic retinopathy may indicate an increased risk of cardiovascular disease in patients with type 1 diabetes—a nested case-control study in Brazil. Front 7. Cardoso CRL, Salles GC, Leite NC, Salles GF. Prognostic impact of carotid Endocrinol. (2019) 10:689. doi: 10.3389/fendo.2019.00689 intima-media thickness and carotid plaques on the development of micro- and macrovascular complications in individuals with type 2 diabetes: the Rio de Janeiro 25. Guo VY, Cao B, Wu X, Lee JJW, Zee BCY. Prospective association between type 2 diabetes cohort study. Cardiovasc Diabetol. (2019) 18:2. doi: 10.1186/s12933- diabetic retinopathy and cardiovascular disease - a systematic review and meta- 019-0809-1 analysis of cohort studies. J Stroke Cerebrovasc Dis. (2016) 25:1688–95. doi: 10.1016/ j.jstrokecerebrovasdis.2016.03.009 8. Vigili de Kreutzenberg S, Coracina A, Volpi A, Fadini GP, Frigo AC, Guarneri G, et al. Microangiopathy is independently associated with presence, severity and 26. Kawasaki R, Tanaka S, Tanaka S, Abe S, Sone H, Yokote K, et al. Risk of composition of carotid atherosclerosis in type 2 diabetes. Nutr Metab Cardiovasc cardiovascular diseases is increased even with mild diabetic retinopathy: the Japan Dis. (2011) 21:286–93. doi: 10.1016/j.numecd.2009.10.003 diabetes complications study. Ophthalmology. (2013) 120:574–82. doi: 10.1016/j. ophtha.2012.08.029 9. Zhang C, Wang S, Li M, Wu Y. Association between atherosclerosis and diabetic retinopathy in chinese patients with type 2 diabetes mellitus. Diabetes 27. Young-Hak K, Myeong-Ki H, Jong-Min S, Ki-Hoon H, Duk-Hyun K, Jae- Metab Syndr Obes Targets Ther. (2020) 13:1911–20. doi: 10.2147/DMSO.S246497 Kwan S, et al. Diabetic retinopathy as a predictor of late clinical events following percutaneous coronary intervention. J Invasive Cardiol. (2002) 14:599–602. 10. Alonso N, Traveset A, Rubinat E, Ortega E, Alcubierre N, Sanahuja J, et al. Type 2 diabetes-associated carotid plaque burden is increased in patients with 28. Klein BEK, Klein R, McBride PE, Cruickshanks KJ, Palta M, Knudtson MD, retinopathy compared to those without retinopathy. Cardiovasc Diabetol. (2015) et al. Cardiovascular disease, mortality, and retinal microvascular characteristics in 14:33. doi: 10.1186/s12933-015-0196-1 type 1 diabetes: wisconsin epidemiologic study of diabetic retinopathy. Arch Intern Med. (2004) 164:1917–24. doi: 10.1001/archinte.164.17.1917 11. Brownrigg JRW, Hughes CO, Burleigh D, Karthikesalingam A, Patterson BO, Holt PJ, et al. Microvascular disease and risk of cardiovascular events among 29. Gubitosi-Klug R, Gao X, Pop-Busui R, de Boer IH, White N, Aiello LP, et al. individuals with type 2 diabetes: a population-level cohort study. Lancet Diabetes Associations of microvascular complications with the risk of cardiovascular disease Endocrinol. (2016) 4:588–97. doi: 10.1016/S2213-8587(16)30057-2 in type 1 diabetes. Diabetes Care. (2021) 44:1499–505. doi: 10.2337/dc20-3104 30. Xu XH, Sun B, Zhong S, Wei DD, Hong Z, Dong AQ. Diabetic retinopathy 12. Drinkwater JJ, Davis TME, Hellbusch V, Turner AW, Bruce DG, Davis WA. predicts cardiovascular mortality in diabetes: a meta-analysis. BMC Cardiovasc Retinopathy predicts stroke but not myocardial infarction in type 2 diabetes: the Disord. (2020) 20:478. doi: 10.1186/S12872-020-01763-Z fremantle diabetes study phase II. Cardiovasc Diabetol. (2020) 19:43. doi: 10.1186/ s12933-020-01018-3 31. Jeevarethinam A, Venuraju S, Dumo A, Ruano S, Mehta VS, Rosenthal M, et al. Relationship between carotid atherosclerosis and coronary artery calcification 13. Mohammedi K, Woodward M, Marre M, Colagiuri S, Cooper M, Harrap S, in asymptomatic diabetic patients: a prospective multicenter study. Clin Cardiol. et al. Comparative effects of microvascular and macrovascular disease on the risk of (2017) 40:752–8. doi: 10.1002/CLC.22727 major outcomes in patients with type 2 diabetes. Cardiovasc Diabetol. (2017) 16:95. doi: 10.1186/s12933-017-0574-y 32. Irie Y, Katakami N, Kaneto H, Takahara M, Nishio M, Kasami R, et al. The utility of ultrasonic tissue characterization of carotid plaque in the prediction 14. Seferovic JP, Bentley-Lewis R, Claggett B, Diaz R, Gerstein HC, Køber LV, of cardiovascular events in diabetic patients. Atherosclerosis. (2013) 230:399–405. et al. Retinopathy, neuropathy, and subsequent cardiovascular events in patients doi: 10.1016/j.atherosclerosis.2013.08.015 with type 2 diabetes and acute coronary syndrome in the elixa: the importance of disease duration. J Diabetes Res. (2018) 2018:1631263. doi: 10.1155/2018/1631263 33. Saeed A, Ballantyne CM. Assessing cardiovascular risk and testing in type 2 diabetes. Curr Cardiol Rep. (2017) 19:19. doi: 10.1007/s11886-017-0831-4 15. Verma S, Bain SC, Honoré JB, Mann JFE, Nauck MA, Pratley RE, et al. Impact of microvascular disease on cardiovascular outcomes in type 2 diabetes: results 34. Catalan M, Herreras Z, Pinyol M, Sala-Vila A, Amor AJ, de Groot E, et al. from the leader and sustain 6 clinical trials. Diabetes, Obes Metab. (2020) 22:2193–8. Prevalence by sex of preclinical carotid atherosclerosis in newly diagnosed type 2 doi: 10.1111/dom.14140 diabetes. Nutr Metab Cardiovasc Dis. (2015) 25:742–8. doi: 10.1016/j.numecd.2015. 04.009 16. Zhu XR, Zhang YP, Bai L, Zhang XL, Zhou JB, Yang JK. Prediction of risk of diabetic retinopathy for all-cause mortality, stroke and heart failure evidence 35. Levey A, Bosch J, Lewis J, Greene T, Rogers N, Roth D. A more accurate from epidemiological observational studies. Medicine. (2017) 96:e5894. doi: 10. method to estimate glomerular filtration rate from serum creatinine: a new 1097/MD.0000000000005894 prediction equation. Modification of diet in renal disease study group. Ann Intern Med. (1999) 130:461–70. doi: 10.7326/0003-4819-130-6-199903160-00002 17. Yamada T, Itoi T, Kiuchi Y, Nemoto M, Yamashita S. Proliferative diabetic retinopathy is a predictor of coronary artery disease in Japanese patients with 36. Wilkinson CP, Ferris FL, Klein RE, Lee PP, Agardh CD, Davis M, et al. type 2 diabetes. Diabetes Res Clin Pract. (2012) 96:e4–6. doi: 10.1016/j.diabres. Proposed international clinical diabetic retinopathy and diabetic macular edema 2011.12.007 disease severity scales. Ophthalmology. (2003) 110:1677–82. doi: 10.1016/S0161- 6420(03)00475-5 18. Zhou JB, Zhu XR, Zhao W, Yin L, Li HB, Qi L, et al. Prediction of proliferative diabetic retinopathy to asymptomatic obstructive coronary artery disease in chinese 37. Touboul P-J, Hennerici MG, Meairs S, Adams H, Amarenco P, Bornstein type 2 diabetes individuals: an exploratory study. Metab Syndr Relat Disord. (2019) N, et al. Mannheim carotid intima-media thickness consensus (2004–2006). 17:367–73. doi: 10.1089/met.2018.0140 Cerebrovasc Dis. (2007) 23:75–80. doi: 10.1159/000097034 Frontiers in Cardiovascular Medicine 10 frontiersin.org Castelblanco et al. 10.3389/fcvm.2022.945421 38. Couronné R, Probst P, Boulesteix AL. Random forest versus logistic 42. Rosolova H, Pelikanova T, Motovska Z. ESC guidelines on diabetes, pre- regression: a large-scale benchmark experiment. BMC Bioinform. (2018) 19:270. diabetes, and cardiovascular diseases developed in collaboration with EASD. Summary of the document prepared by the czech society of cardiology. Eur Heart 39. Spirtes P, Glymour C, Scheines R. Causation, Prediction, and Search. 3rd ed. J. (2013) 34:3035–87. doi: 10.1016/j.crvasa.2014.01.007 Cambridge, MA: The MIT Press (2001). 43. Wilson PWF, Polonsky TS, Miedema MD, Khera A, 40. R Core Team. R: A language and environment for statistical computing. Kosinski AS, Kuvin JT. Systematic Review for the 2018 Vienna, Austria: R Foundation for Statistical Computing (2020). AHA/ACC/AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA 41. Carbonell M, Castelblanco E, Valldeperas X, Betriu A, Traveset A, Granado- guideline on the management of blood cholesterol: a report of the American Casas M, et al. Diabetic retinopathy is associated with the presence and burden of college of cardiology/American heart association task force on clinical practice subclinical carotid atherosclerosis in type 1 diabetes. Cardiovasc Diabetol. (2018) guidelines. J Am Coll Cardiol. (2019) 73:3210–27. doi: 10.1016/j.jacc.2018.1 17:66. doi: 10.1186/s12933-018-0706-z 1.004 Frontiers in Cardiovascular Medicine 11 frontiersin.org