The Association of Severe Diabetic Retinopathy With Cardiovascular Outcomes in Long-standing Type 1 Diabetes: A Longitudinal Follow-up

Cardiovascular disease (CVD) is the leading cause of death in patients with long-standing type 1 diabetes (T1D) (1). Diabetic kidney disease (DKD) is well recognized as one of the most important predictors of CVD and premature mortality (2). However, the independent association of diabetic retinopathy, another prevalent microvascular disease, with CVD is less clear.

The retina offers a unique feature of direct visualization of the vasculature. Attempts have been made to investigate the association between the retinal microcirculation and macrovascular disease in various populations. Changes in the retinal arteriolar and venular caliber were associated with increased risk of stroke (3–5) and cardiovascular mortality in subjects without diabetes (6,7). Similarly, diabetic or other retinal vascular disease has been suggested to predict peripheral artery disease (PAD) and cardiovascular mortality in type 2 diabetes (8–11). We have previously shown a relationship among diabetic retinopathy, stroke, and mortality in T1D (2,12). An association between diabetic eye disease and coronary heart disease (CHD) was also reported in different populations with T1D (13–16). In the majority of these studies, it has not been possible to dissect the role of diabetic retinopathy from the established risk factors, such as DKD in particular, or to analyze the data in the context of different cardiovascular outcomes (17,18). However, the recent cross-sectional Joslin 50-Year Medalist Study together with a subsample of individuals taking part in the FinnDiane study suggested that proliferative diabetic retinopathy (PDR) without DKD increased the CVD risk (19).

Therefore, we aimed to examine whether severe diabetic retinopathy (SDR) per se, independent of DKD, predicts CVD and its different subtypes in a well-characterized large cohort of individuals with duration of >30 years of T1D. Furthermore, due to the paucity of data on CVD outcomes in patients with >30 years of diabetes duration (20), we also explored the difference in relative risk of CVD in T1D compared with a matched population without diabetes. Finally, we examined whether the addition of SDR as a risk factor improves the CVD risk prediction in individuals with T1D.

This study is part of the ongoing Finnish Diabetic Nephropathy (FinnDiane) Study, which is a nationwide, prospective, multicenter study with the aim to identify risk factors for T1D complications (12). Data collection started in 1997 and now involves all 5 university hospitals, all 16 central hospitals, the majority of the regional hospitals, and several primary health care centers. For this study, we included all patients (n = 1,683) with at least 30 years of T1D duration at baseline or a follow-up visit and complete information available on renal and retinal status. Two to three control individuals for each patient were selected from the Population Register Center matched for sex, age, and the place of residence in the year of diagnosis of diabetes. After exclusion of people with diabetes (n = 636), 4,016 control subjects remained in the study. All patients signed a written informed consent. The study protocol was approved by the ethics committee of the Helsinki and Uusimaa Hospital District and separately by each study center, and the study was performed in accordance with the Declaration of Helsinki.

T1D was defined as diabetes diagnosed before 40 years of age and permanent insulin therapy initiated within 1 year of the diabetes diagnosis. Data on sex were self-reported, and all patients were Caucasian. Hypertension was defined as either systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg, based on the average of two measurements, or any use of antihypertensive medication. Data on medication were collected from the patient questionnaire and then confirmed by linking the FinnDiane data to the nationwide Drug Prescription Register (Supplementary Table 1B). Smoking was defined as previous or current smoking of at least 1 cigarette/day for at least 1 year.

Serum samples were analyzed for HbA_1c, lipids, and creatinine. The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology Collaboration formula (21). DKD was defined as eGFR <60 mL/min/1.73 m². In addition, albuminuria status was determined based on measurements of urinary albumin excretion rate (UAER) from at least two timed overnight (g/min) or 24-h (mg/24 h) urine collections as reported before (12). Normoalbuminuria was defined when in at least two out of three urinary samples’ UAER was <20 μg/min or <30 mg/24 h; microalbuminuria, when UAER was ≥20 and <200 μg/min or ≥30 and <300 mg/24 h; or macroalbuminuria, when UAER was ≥200 μg/min or ≥300 mg/24 h.

In this study, the retinopathy status was based on a history of laser photocoagulation. Severe diabetic retinopathy (SDR) was defined as initiation of laser treatment due to severe non-PDR, PDR, or diabetic maculopathy. The underlying cause for laser treatment was PDR in the majority (>80%) of patients, and the remaining patients received laser photocoagulation because of macular edema or severe non-PDR (22). Of note, most of the patients that received laser treatment primarily for macular edema later received this treatment also for PDR. By using laser treatment as the definition of SDR (i.e., information that was available for all study patients), we were able to substantially enlarge our study population compared with our previous collaborative study with the Joslin Medalist Study that was based on the Early Treatment Diabetic Retinopathy Study (ETDRS) classification (19).

Cardiovascular outcomes were retrieved from the Finnish Care Register for Health Care (nonfatal cases) and the Cause of Death Register (fatal cases) at the end of 2015. CVD was defined as CHD, any stroke (ischemic and hemorrhagic), or PAD, whichever occurred first. CHD included the presence of myocardial infarction or coronary revascularization (percutaneous coronary intervention or coronary artery bypass graft). PAD was defined as a nontraumatic lower extremity amputation at any level or open revascularization of lower extremity. The risk of ischemic stroke was also evaluated separately. Specific codes used are listed in Supplementary Table 1B.

At baseline (during 1995–2015), patients were divided into four groups according to the presence of DKD and/or SDR. Differences among groups were analyzed by ANOVA for normally distributed and Kruskal-Wallis test for nonnormally distributed variables. Kaplan-Meier plots were used to illustrate the proportion of patients who endured cardiovascular outcomes.

Follow-up regarding incident CVD events started at baseline and ended at CVD event, death, or the end of 2015. For these analyses, previous and thus prevalent CVD events (all CVD, n = 368; CHD, n = 194; PAD, n = 150; and stroke, n = 107) were excluded.

Cox regression analysis with forward stepwise entry was performed to determine factors independently associated with CVD, separately in patients with or without DKD and with or without SDR at baseline (+DKD/+SDR, +DKD/−SDR, −DKD/+SDR, and −DKD/−SDR). Possible development of DKD or SDR during follow-up was not taken into account. Kidney status (DKD) in the main analyses was based on eGFR (<60 mL/min/1.73 m² or ≥60 mL/min/1.73 m²) as in our previous study (19), with additional analyses performed according to albuminuria status (Supplementary Tables 4–6).

Variables for the model were chosen based on significant univariable associations. We included triglyceride-to-HDL cholesterol ratio (TG-HDL ratio) rather than LDL cholesterol because the former was a direct measurement. Smoking was not associated with CVD in this cohort, but because it is a well-established cardiovascular risk factor (23), it was included in the model. Thus, variables included were age at diabetes diagnosis, diabetes duration, sex, HbA_1c, hypertension, TG-HDL ratio, waist-to-hip ratio (WHR), history of hospitalization due to severe hypoglycemia, and smoking. Cox proportional hazards analysis was performed to determine the risk related to a pooled CVD outcome as well as to the specific CVD outcomes (CHD, ischemic stroke, and PAD). Moreover, the role of SDR in shaping cardiovascular risk was ascertained in dependence of different albuminuria strata. In order to get a comprehensive view of the occurrence of CVD along the duration of diabetes and attained age, age- and duration-specific incidence rates were calculated for CVD from the diagnosis of diabetes.

Excess CVD morbidity in the patients with T1D compared with that in the control subjects without diabetes was determined by calculating standardized incidence ratios (SIRs) as ratios of observed and expected numbers of case subjects. The expected numbers were derived by multiplying the number of person-years at risk by sex-, age-, and period-specific incidence rates observed in the control individuals without diabetes. Analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC).

After exclusion of overlapping events, there were altogether 784 CVD events, 368 prevalent CVD events, and 416 incident CVD during follow-up. However, by counting each CVD event separately, there were 354 incident CHD, 205 incident PAD, and 187 incident strokes. The overall follow-up time for the incident CVD events was 12,872 person-years, with a median follow-up of 9.5 (interquartile range 4.1–14.1) years; for the CHD events, 11,824 person-years, median 10.7 (5.8–15.0) years; for stroke, 16,121 person-years, median 10.4 (5.0–14.7) years; and for the PAD events, 15,407 person-years, median 10.4 (4.9–14.8).

Baseline characteristics of the four DKD/SDR status groups are summarized in Table 1. The group with DKD but without SDR was the smallest (n = 74). It had the longest diabetes duration, the highest BMI and HbA_1c, and the largest proportion of smokers (P < 0.001). On the contrary, the group without both DKD and SDR (n = 527) had the latest diabetes onset, the lowest HbA_1c, TG-HDL ratio, WHR, and blood pressure, and the lowest number of smokers (P < 0.001). Notably, 30% of the patients did not experience either SDR or DKD nor any cardiovascular event. They were more likely females, had later diabetes onset and lower blood pressure, HbA_1c, TG-HDL ratio, and WHR, and were less likely to be previous smokers and to have had a history of hospitalization due to severe hypoglycemia compared with patients with at least one listed complication present (Supplementary Table 2). Patient characteristics with CVD in relation to the ones without CVD and depending on the DKD/SDR status are presented in Supplementary Table 3. Significant CVD risk predictors were diabetes duration, age at diabetes onset, HbA_1c, blood pressure, dyslipidemia, history of hypoglycemia, and decreased eGFR (Table 2).

In univariable analysis, SDR was a predictor of incident cardiovascular events in patients without DKD (hazard ratio [HR] 1.61 [95% CI 1.23–2.10]). The 15-year cumulative incidence of any CVD in patients with and without SDR was 36.8% (95% CI 33.4–40.1) and 27.3% (23.3–31.0), respectively (P = 0.0004 for log-rank test) (Fig. 1A). The results were similar when, in additional analyses, the kidney status was based on albuminuria; the 15-year cumulative incidence of any CVD in patients without albuminuria but with SRD was 29.0% (95% CI 24.0–33.8) and in those without both conditions 23.2% (95% CI 18.9–27.2) (HR 1.54 [95% CI 1.12–2.12]) (Fig. 1B).

In the multivariable model, the association remained significant after adjustment for known CVD risk factors (Table 2). Furthermore, SDR was also independently associated with CHD and PAD. The risk of ischemic stroke was doubled in the presence of SDR; however, after adjustment for traditional risk factors, the association was no longer significant (P = 0.08) (Table 2). The SIR without SDR and without DKD was 3.5 (95% CI 2.9–4.1). However, it was higher, 5.6 (4.9–6.4), if SDR was present.

The risk of CVD was significantly higher in patients with DKD. The 15-year cumulative incidence of any CVD was 71.5% (95% CI 69.3–73.4) and 58.2% (50.6–64.6) in patients with and without concomitant SDR, respectively (Fig. 1A). The corresponding results were 59.9% (95% CI 57.6–62.1) and 49.6% (43.7–54.9) in patients with albuminuria and SDR and in patients with albuminuria but without SDR, respectively (Fig. 1B).

Compared with patients without DKD and without SDR, the overall risk of CVD and CHD was two to three times higher in the presence of DKD. Similarly, it was two- to fivefold increased for ischemic stroke and PAD (Table 2). The risk of CVD was also higher in patients with adjunct SDR on top of DKD compared with those with DKD but without SDR (HR 1.46 [95% CI 1.11–1.92]). The SIR for CVD between those with T1D and DKD but without SDR and control individuals without diabetes was 11.5 (95% CI 8.5–15.1), but concomitant SDR increased the SIR to 16.1 (14.5–17.7). The associations of SDR with CVD, adjusted for different stages of albuminuria, are reported in Supplementary Tables 4–6.

Patients without DKD and SDR at baseline had 4.0-fold (95% CI 3.3–4.7) increased risk of CVD compared with control subjects without diabetes up to 70 years of age (Fig. 2A). Intriguingly, after this age, the CVD incidence was similar to that in the matched control subjects (SIR 0.9 [95% CI 0.3–1.9]) in this subgroup of patients with diabetes. However, in patients without DKD but with SDR, the CVD risk was still increased after the patients had reached 70 years of age (SIR 3.4 [95% CI 1.8–6.2]) (Fig. 2A). Of note, in patients with both DKD and SDR, the CVD burden was high already at young ages. Figure 2B shows the CVD risk according to duration of diabetes. The enhancing effect of SDR without DKD was most apparent after long duration of diabetes.

This study highlights the role of SDR on a complete range of CVD outcomes in a large sample of patients with long-standing T1D and longitudinal follow-up. We showed that SDR alone, without concomitant DKD, increases the risk of macrovascular disease, independently of the traditional risk factors. The risk is further increased in case of accompanying DKD, especially if SDR is present together with DKD. Findings from this large and well-characterized cohort of patients have a direct impact on clinical practice, emphasizing the importance of regular screening for SDR in individuals with T1D and intensive multifactorial interventions for CVD prevention throughout their life span.

This study also confirms and complements previous data on the continuum of diabetic vascular disease, by which microvascular and macrovascular disease do not seem to be separate diseases, but rather interconnected (10,12,18). The link is most obvious for DKD, which clearly emerges as a major predictor of cardiovascular morbidity and mortality (2,24,25). The association of SDR with CVD is less clear. However, our recent cross-sectional study with the Joslin Medalist Study showed that the CVD risk was in fact increased in patients with SDR on top of DKD compared with DKD alone (19). In the present longitudinal study, we were able to extend those results also to show that SDR alone, without DKD and after the adjustment for other traditional risk factors, increases CVD risk substantially. SDR further increases CVD risk in case DKD is present as well. In addition, the role of SDR as an independent CVD risk predictor is also supported by our data using albuminuria as a marker of DKD. This is important because albuminuria is a known predictor of diabetic retinopathy progression (26) as well as a recognized biomarker for CVD. Of note, the association of SDR with CVD reached statistical significance in the adjusted analysis if SDR was present in patients with no or incipient albuminuria (normoalbuminuria and microalbuminuria), but not when it was present in patients with normoalbuminuria only. We hypothesize that this is because SDR together with the more advanced levels of albuminuria better reflect definite disease phenotypes than their less advanced forms. In this study, due to the low number of patients with nonalbuminuric DKD (n = 44), we could not analyze whether albuminuric DKD further strengthens the association of SDR with CVD compared with nonalbuminuric DKD.

Previous studies on T1D such as the Wisconsin study showed that patients with diabetic retinopathy had increased risk of cardiovascular mortality, CHD, and stroke, a risk that was, however, not independent from diabetic nephropathy (27). In the Pittsburgh cohort, an association between the retinal arteriolar diameter and CHD was present only in women, yet adjustments were made only for traditional risk factors but not for diabetic nephropathy (28). Moreover, the association lost significance after a more stringent criterion for CHD (excluding angina pectoris symptoms and electrocardiogram changes) was considered. Similarly, there was either no link of retinopathy with CVD, or it disappeared after adjustment for diabetic nephropathy in other studies (13–16). In our study, we observed a significant, independent effect of SDR on the CVD incidence overall and also separately for CHD and PAD in patients without DKD. One possible reason for this discordant result may be the lack of enough power in previous studies, particularly with regard to CVD events. Another reason may be the long diabetes duration, because our cohort has the longest T1D duration described, and it also includes longitudinal data. Diabetes duration may also place our current findings in the context of the previous report that mortality in T1D is increased only in the presence of DKD (2). The patient cohort in the current study has survived long years despite suffering from T1D and may thus have a protective factor against glucotoxicity on the kidneys. These patients are therefore ideal to study associations between SDR alone and CVD, hence an association with SDR and CVD became evident.

A novel finding is that, independently of any signs of DKD, the risk of PAD is increased twofold in the presence of SDR. Although this association has recently been highlighted in individuals with type 2 diabetes (10,29), the data in T1D are scarce (16,30). Notably, the previous studies mostly lack adjustments for DKD, the major predictor of mortality in patients with shorter diabetes duration. Both complications, besides sharing some conventional cardiovascular risk factors, may be linked by additional pathological processes involving changes in the microvasculature in both the retina and the vasa vasorum of the conductance vessels (31).

Specifically, neovascularization, mediated by vascular endothelial growth factor or hypoxia-inducible factor-1α in the ischemic retina, operates also in the vasa vasorum of plaques in large vessels (32) and was shown to play a role in accelerated plaque progression and plaque instability (33). Also, the link may be neurovascular communication. The combined effect of neuronal and vascular damage on the lower extremity amputation rate is well known; however, more recently, the role of neuroprotection in PDR treatment is being recognized (34).

Intriguingly, we found no association between SDR and ischemic stroke in patients without DKD. As the retinal microvasculature shares embryologic, anatomic, and regulatory characteristics with that of the cerebral circulation (35,36), one would expect that retinal changes parallel those in the cerebral vasculature, with the association of SDR with stroke being particularly strong. Nevertheless, such an association has been seen in individuals with type 2 diabetes (37,38), in the general population (3,5), and even in individuals with T1D, as previously shown by us (12). One possible explanation for this discrepancy could again be that the patients analyzed in the current study had a substantially longer diabetes duration, reflecting a changing picture of vascular risk factors at different stages of diabetes duration. Curiously, however, in our previous study, there was no link found between the lacunar stoke proportion and SDR, even though lacunar stroke is regarded as a microvascular cerebral disease (39). The underlying explanation may be that rather than the typical signs of SDR, more subtle changes in the retinal vascular architecture, such as changes in the arterio-venular ratio and microvascular branching are associated with cerebral disease, highlighting the role of more detailed studies of retinal microvasculature for CVD risk determination (3,5) in individuals with T1D.

Surprisingly, smoking was not a significant predictor of CVD in this cohort. We believe that this is due to survival benefit of nonsmokers and the negative effects of smoking being diluted by other risk factors, particularly DKD.

In contrast to the appealing suggestion that surviving long-standing T1D without apparent chronic microvascular complications may indicate an exposure of those patients to a yet-unidentified protective factor (40), our findings show CVD risk is increased even in patients without SDR and DKD compared with age-matched control subjects without diabetes. Yet, our cohort without microvascular complications, with slightly shorter diabetes duration, also seemed to be protected, but only after reaching their seventies (Fig. 2). This finding underscores the need for lifelong aggressive CVD protection, even in patients without apparent diabetic chronic complications.

The strengths of this study include the large number of participants, collected from all over Finland at all levels of health care, with exceptionally long diabetes duration and ∼10 years of follow-up. All patients had access to subsidized care (75–100% of costs) and access to established CVD protection therapies, such as statins (although the rate of use was low) and blockers of the renin-angiotensin-aldosterone system.

Our study has also several limitations. The group with DKD but without SDR is rather small. Therefore, the chance to find a significant independent effect of SDR in the presence of DKD is less likely. Yet, from the number of participants in this group, we can conclude that having DKD without SDR is out of the ordinary, making these associations less relevant. Also, CVD data were collected from the national registries instead of the patients’ files. However, when data from the national registries were compared with individual patient hospital records previously (41), no classification errors were found, confirming the Finnish registry data to be a reliable source for CVD outcomes identification. In addition, we present data only on SDR. It would be interesting to investigate whether earlier retinal damage could be informative on CVD risk. Likewise, the possibility of adding data on diabetic neuropathy could potentially provide further insights to the CVD risk as was described in type 2 diabetes (42).

Patients with T1D duration of >30 years face a continuously increased CVD risk that is further increased by the occurrence of advanced PDR. Therefore, by examining the retina, additional insight into individual CVD risk is gained and can guide the clinician to a more tailored approach to CVD prevention. Moreover, our findings suggest that the link between SDR and CVD is at least partially independent of traditional risk factors, and the mechanism behind the phenomenon warrants further research, aiming to find new therapies to alleviate the CVD burden more efficiently.

Acknowledgments. The authors thank all physicians and nurses at each FinnDiane center participating in patient recruitment and characterization. The complete list of physicians and nurses is presented in the Supplementary Data.

Funding. This research was funded by grants from the Folkhälsan Research Foundation, Academy of Finland, Wilhelm and Else Stockmann Foundation, Liv och Hälsa Society, Novo Nordisk Foundation, Diabetes Research Foundation, and Päivikki and Sakari Sohlberg Foundation. D.P.B. was supported by the European Association for the Study of Diabetes Albert Renold Fellowship, generously offered by the European Foundation for the Study of Diabetes.

Duality of Interest. D.P.B. reports receiving lecture honoraria from Boehringer Ingelheim, AstraZeneca, Novo Nordisk, Eli Lilly and Company, Merck, Merck Sharp & Dohme, Krka, and Servier Pharma and being an advisory board member of Boehringer Ingelheim, AstraZeneca, and Novo Nordisk. P.-H.G. reports receiving lecture honoraria from AstraZeneca, Boehringer Ingelheim, Eli Lilly and Company, ELO Water, Genzyme, Merck Sharp & Dohme, and Novartis and being an advisory board member of AbbVie, AstraZeneca, Boehringer Ingelheim, Eli Lilly and Company, Janssen, Medscape, Merck Sharp & Dohme, Novartis, Novo Nordisk, and Sanofi. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.P.B. participated in the study conception, literature search, data analysis, data interpretation, and drafting of the manuscript. V.H. was responsible for the study design, data acquisition, statistical analyses, data interpretation, and drafting of the manuscript. D.G. and G.K. participated in design of the study, data interpretation, and critical revision of the manuscript for important intellectual content. M.K. and C.F. contributed to data acquisition, data interpretation, and critical revision of the article. P.-H.G. is the principal investigator of the study and participated in the study conception, data interpretation, and critical revision of the manuscript for important intellectual content. P.-H.G. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.