Retinopathy, retinal measurements improve predictive value of existing CVD risk factors
Incorporating retinal microvascular parameters and retinopathy substantially improves the predictive value of existing risk factors for cardiovascular disease (CVD) in individuals with diabetes, reports a new prospective cohort study.
“The addition of retinal and serum biomarkers markers significantly improved the predictive value of models compared to traditional risk factors alone, with the addition of retinal parameters proving to be stronger predictors of CVD compared with serum biomarkers of [high-sensitivity C-reactive protein (hsCRP)] and [estimated glomerular filtration rate (eGFR)],” the investigators wrote.
Nonfasting blood samples from 709 diabetes patients enrolled in the Singapore Malay Eye Study (SiMES) were used to create the CVD prediction model which was then verified against 425 fasting blood samples from diabetics enrolled in the Singapore Prospective Study Program (SP2).
Retinopathy and retinal microvasculature, presented as central retinal artery (CRAE) and central retinal vein (CRVE) equivalents, were measured using digital fundus photography. These were then included and tested against established risk factors such as serum lipid biomarkers, eGFR, smoking, BMI and blood pressure.
Overall, the SiMES cohort had significantly higher smokers (p=0.05), hypertensives (p<0.001), systolic blood pressure (SBP; p<0.001), total and low-density lipoprotein cholesterol (LDL-C; p<0.001 for both), HbA1c (p<0.001), eGFR (p<0.001) and percentage of females (p=0.004) compared with the SP2 cohort. There were no significant differences in terms of CRAE, CRVE and hs-CRP. [Sci Rep 2017;7:41492]
Over a mean follow-up period of 6.1 (0.2 to 7.1) years, 86 CVD cases (12.1 percent) were reported in the SiMES cohort. Of these, 16 were of mixed origins, 19 were deaths, 22 were strokes and 29 were myocardial infarctions (MI).
On the other hand, 31 CVD cases (7.3 percent) were reported in the 5.78 (0.29 to 7.18)-year follow-up of the SP2 cohort. Deaths were reported in 4 cases, strokes and mixed aetiology in 7 each, and MIs in 13. While there were significantly more CVD events in the SiMES cohort (p=0.01), there were no statistically significant differences in terms of MI (p=0.298) or stroke (p=0.187).
Including hs-CRP (hazard ratio [HR], 1.41; 95 percent CI, 1.10 to 1.79) and eGFR (HR, 1.02; 1.01 to 1.03), both significantly correlated with CVD risk, moderately improving the reference model with a C statistic of 0.747 (p=0.0556). Adding only retinal measures marginally improved the C statistic further to 0.751 (p=0.0542).
However, including hs-CRP (HR, 1.43; 1.12 to 1.82), eGFR (HR, 1.01; 1.01 to 1.03) and presence of retinopathy (HR, 1.94; 1.17 to 3.21), with all tied to CVD risk, significantly improved the discriminative value of the model resulting in a C statistic of 0.751 (p=0.011).
The final model included all factors including hs-CRP (HR, 1.41; 1.10 to 1.81), eGFR (HR, 1.01; 1.01 to 1.02), retinopathy (HR, 2.04; 1.27 to 3.28) and CRAE (HR, 1.35; 1.04 to 1.83), all of which correlated significantly with CVD risk. This resulted in a statistically improved C statistic of 0.774 (p=0.003) compared to the reference model.
Validation of the final model against the SP2 cohort showed generally similar trends and magnitudes of the HRs except for retinopathy (HR, 0.98; 0.42 to 2.28), which lost its significance and CRVE (HR, 2.15; 1.38 to 3.32), which gained significance. The C statistic improved further from 0.763 to 0.813 (p=0.045) compared to the reference model in SP2.
Existing literature has established the correlative relationship between changes in retinal vasculature and CVD-related and all-cause mortality, thus potentially implicating them in alterations in the circulation of larger systems, the authors wrote.
Similarly, changes in the microcirculatory systems in the context of diabetes have been extensively studied. The current literature “suggests that retinopathy and retinal microvascular changes are sensitive markers of microvascular dysfunction related to CVD in patients with diabetes,” they explained.
Thus, the current study synthesized these bodies of knowledge, testing for the association between CVD and retinal measurements in diabetics, and used them in the construction of externally-validated and generalized predictive models.
“[W]e demonstrated that an assessment of retinopathy and retinal microvascular measures captured from retinal photographs provides prognostic information on risk of CVD and significantly improves CVD risk stratification when incorporated into established risk models in persons with diabetes,” the investigators said.
“We also demonstrated that a ‘multiple marker’ approach might be better than traditional risk scores for CVD prediction in patients with diabetes,” they added.