Diabetic retinopathy as a marker of cognitive dysfunction and depression: a clinical and epidemiological approach

Introduction and aim
Diabetes mellitus is a global health concern affecting an estimated number of 537 million people,with diabetic retinopathy (DR) being the most common complication. In recent years retinal neurodegeneration has been recognized as an early occurrence in DR. Given that the eye and brain sharesimilar embryologic origin, anatomical features and physiological properties, observations suggestthat the retina provides a unique “window” to the brain. Furthermore, diabetes has been linked toincreased risk of systemic neurodegenerative diseases such as Alzheimer’s disease (AD) and depression, but less is known regarding the association with DR. In addition non-invasive retinal parameters have been associated with systemic neurodegenerative diseases, but only a few studieshave examined these associations in a high-risk population of individuals with diabetes.Thus, with a register-based and clinical approach we aimed to investigate DR as a marker of AD(Paper I) and depression (Paper II) utilizing real world data, and explore in a clinical setting, whethernon-invasive retinal parameters are associated with mild cognitive impairment (MCI) (Paper III), apre-state of dementia, and depression (Paper IV) in individuals with type 2 diabetes.


Methods
In the register-based approach, we utilized various Danish nationwide registers as data sources. Thisincluded the Danish Registry of Diabetic Retinopathy and Maculopathy (DiaBase), the Danish National Patient Registry (DNPR), the Danish National Prescription Registry (NPR), the Danish CivilRegistration System, and the Register of Laboratory Results for Research. Level of DR was extractedfrom Diabase and diabetes type was defined by combining International Classification of Diseases(ICD) 10th revision codes of diabetes from the DNPR and Anatomical Therapeutic Chemical Classification codes on glucose lowering treatment from the NPR. We defined AD by ICD-10 codes F00*(Dementia in Alzheimer’s disease) or G30* (Alzheimer’s disease) given once by either a geriatrician,psychiatrist or neurologist (Paper I) whereas depression was defined by ICD-10 codes F32 (depressive episode), F33 (recurrent depressive disorder) and F34.1 (dysthymia) (Paper II). We appliedmultivariable logistic regression and Cox regression models and reported odds ratios (OR) and hazard ratios (HR) with 95% confidence intervals.

In the clinical approach, we performed a cross-sectional study of individuals with type 2 diabetesabove the age of 65 years and duration of diabetes of more than 5 years. To evaluate cognitivefunction and depression, we conducted thorough neuropsychological assessments and self-administered depression questionnaires. We also performed retinal examinations including color fundusphotography (CFP), retinal oximetry, optical coherence tomography (OCT) and OCT-angiography(OCT-A). DR was assessed by a certified grader with the level of severity classified according to theInternational Clinical Diabetic Retinopathy severity scale. Retinal parameters of interest were theretinal vessel oxygen saturation, retinal vessel width, tortuosity, density, fractal dimension, macular retinal nerve fiber layer (mRNFL), macular ganglion cell layer (mGCL), size of foveal vascular zoneand size of non-perfusion area.


Main results and conclusion
In Paper I, we included 134,127 individuals with diabetes above 60 years of age and 651,936 ageand sex matched individuals without diabetes. At baseline, individuals with diabetes were less likelyto have AD (OR 0.63 [95% CI 0.59;0.68]) compared to age and sex matched individuals. Duringfollow-up individuals with diabetes without DR had a decreased risk of incident AD (HR 0.87 [95% CI0.81;0.93]), but individuals with diabetes and DR had a 24% higher risk of incident AD compared toage and sex matched individuals (HR 1.24 [95% CI 1.08;1.43]).

In Paper II, we included 240,893 individuals with type 2 diabetes. Those with DR were also less likelyto have a history of diagnosed depression compared to those without DR (OR 0.79 [95% CI0.74;0.83]). During follow-up, those with DR had a significantly higher risk of incident diagnosed depression, but after adjustment of confounders, DR was not independently associated with increasedrisk of diagnosed depression (HR 1.03 [95% CI 0.94;1.13]). In reverse we found a lower risk of incident DR in those with a history of diagnosed depression (HR 0.90 [95% CI 0.84-0.95]), but the association did not persist in a sensitivity analysis, where we restricted the diagnosis of depression to begiven within two years prior of the screening episode (HR 1.07 [95% CI 0.91;1.26]).

In the clinical part of the study, we included 134 individuals with type 2 diabetes. We did not find adifference in the overall retinal oxygen saturation, but in a sub-analysis we found reduced retinalvenular oxygen saturation (61.0 vs. 56.3%, p=0.024) in the upper nasal quadrant to be associatedwith MCI. Furthermore, we found that non-invasive retinal neural parameters defined as thinnermRNFL (22.3 vs. 24.1 µm, p=0.010) and mGCL (45.7 vs. 48.4µm, p=0.030) in the superior regionwere associated with MCI (Paper III). We did not find any association with retinal microvascular parameters and MCI. When evaluating the association between depression and retinal parameters wefound an association between retinal arterial oxygen saturation in the lower temporal quadrant anddepression (per 1% increment, OR 1.07 [95% CI 1.01;1.14]). We found no association between neurovascular retinal parameters and depression.

This dissertation reports an association between DR and systemic neurodegenerative diseases,which increases the risk of AD (Paper I), but not independently with diagnosed depression (Paper II).On the contrary, diagnosed depression was associated with lower risk of incident DR, but the temporality between diagnosis of depression and screening of DR was important. In a population withtype 2 diabetes, we found retinal neurodegeneration, given as lower mRNFL and mGCL, to be associated with MCI (Paper III). In addition, we found some evidence of an association between retinaloxygen saturation and depression but not with neurovascular parameters (Paper IV). Non-invasiveretinal parameters may in the future be utilized in diagnostic, prognostic or management of both MCIand depression in individuals with type 2 diabetes, but more extensive work is necessary within thisarea before findings as ours can be implemented in daily clinical care.