The association between diabetic retinopathy and cognitive impairment in patients with diabetes mellitus: a systematic review and meta-analysis.

The association between diabetic retinopathy and cognitive impairment in patients with diabetes mellitus: a systematic review and meta-analysis

Abstract

Background

Diabetic retinopathy (DR) is a recognized microvascular complication of diabetes in which neurovascular and neurodegenerative mechanisms may also contribute. This systematic review evaluated the association between DR and cognitive impairment or dementia, including Alzheimer’s disease and vascular dementia, using evidence from cross-sectional and longitudinal observational studies.

Methods

We searched for relevant articles up to December 2025 from PubMed, Scopus, Web of Science, and Cochrane Library. Two independent reviewers extracted data from the selected studies, including baseline information and outcomes. All quantitative syntheses were conducted using R version 4.4.1.

Results

Fourteen studies with reported study populations totaling 2,871,847 participants were included. Seven studies contributed to the crude dichotomous synthesis, which did not demonstrate a statistically significant association between diabetic retinopathy and cognitive impairment or dementia (OR 1.07, 95% CI 0.94–1.23; p=0.310; I2=52.3%). Seven studies contributed adjusted longitudinal estimates for incident dementia; diabetic retinopathy was associated with a higher incidence of dementia (HR 1.55, 95% CI 1.19–2.02; p=0.001; I2=97.8%; 95% prediction interval 0.65–3.68). Evidence concerning Alzheimer’s disease and vascular dementia remained inconsistent.

Conclusion

Longitudinal evidence suggests that diabetic retinopathy is associated with a higher incidence of dementia, although substantial heterogeneity and the observational evidence base limit confidence in the pooled estimate. Evidence concerning Alzheimer’s disease and vascular dementia remains inconclusive.

Systematic review registration

https://doi.org/10.17605/OSF.IO/YS462, identifier YS462.

Keywords: Alzheimer’s disease, cognitive impairment, dementia, diabetes mellitus, diabetic retinopathy

1.Introduction

Retinopathy, neuropathy, and nephropathy are classically recognized complications of diabetes mellitus (DM) (1). However, there has been a growing body of literature addressing the associated cognitive decline, mild cognitive impairment (MCI), and dementia in patients with diabetes (2), with multiple studies reporting an association between DM and dementia subtypes, Alzheimer’s disease (AD), and vascular dementia (VaD) (3–5). Contemporary syntheses remain equivocal in determining the exact underlying mechanism through which diabetes impacts cognitive functions, with various theories implicating microvascular complications, glycemic strain, systemic cardiometabolic comorbidities, neuroinflammation, or other neurodegenerative processes (3, 6, 7).

While previously considered a microvascular complication of diabetes, evidence suggests that DR is attributable to both microvascular insults and neurodegeneration, such as inner retinal neurodegeneration, manifested through ganglion cell depletion, reactive gliosis, inner retinal thinning, and functional deficits observed via electroretinogram (8). In addition, these neuro-retinal impairments often manifest prior to the emergence of clinically detectable microvascular lesions (9). These findings necessitate a paradigm shift in our understanding of DR, suggesting that the condition is a neurodegenerative ocular disease that initiates during the earliest phases of diabetic pathology (10, 11). Early impairment of the neurovascular unit in the retina of patients with diabetes is evidenced by diminished vasomotor responses to hyperoxic and flickering light stimuli, reflecting a systemic regulatory breakdown rather than isolated vascular or glial damage. These preclinical disturbances are analogous to neurovascular dysfunction seen in cerebral pathologies like cerebrovascular insults and AD (8, 12, 13).

Research indicates that type 2 DM significantly alters both the physical structure and functional activity of the brain (14). Structurally, MRI scans reveal that diabetic individuals often experience brain atrophy, particularly in the hippocampus and temporal lobe, alongside a higher prevalence of small infarcts and hyperintense lesions in the white matter (15). Even in the early stages of cognitive decline, functional imaging shows that diabetes disrupts neural signaling in regions responsible for emotion and thought, such as the frontal lobe and amygdala, while also causing abnormal activity in the visual processing centers (16, 17).

Prior meta-analyses were met with several limitations. For instance, one meta-analysis could not determine whether the association between DR and CI was higher with increasing severity of DR, such as in patients with proliferative DR, due to the mixed nature of the analyzed data. Moreover, inconsistent adjustments for confounding factors across pooled data precluded combining the estimates in a multivariate model (18). In another meta-analysis, the variability in the tools used to assess cognitive function confounded the pooled analysis, with high heterogeneity attributed to multiple factors, including the type of diabetes, disease duration, and other demographic factors (19). Moreover, the underpowered estimates for some subtypes of dementia, such as VaD, prevented firm conclusions. Various sources of bias also existed, including survival bias, diagnostic biases due to self-reported data, and biases due to the reliance on medical records, which skews the results toward treated patients (20, 21). Accordingly, this updated review evaluated the association between DR and cognitive impairment or dementia using a search extending through 2025. Unlike previous reviews, we separated crude dichotomous associations from adjusted longitudinal hazard estimates, verified that population-based studies provided an explicitly DR-specific or diabetes-restricted estimate, incorporated the recently published Lee et al. cohort, and summarized Alzheimer’s disease and vascular dementia findings separately.

2.Methods

Following the guidelines outlined in the Cochrane Handbook for Systematic Reviews of Interventions and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), we conducted a meta-analysis to examine the relationship between DR and CI/dementia in patients with diabetes (22, 23). The review protocol was registered on the Open Science Framework (OSF) and is available at: https://doi.org/10.17605/OSF.IO/YS462.

2.1.Eligibility criteria

Eligibility was defined using the PICOS framework, with diabetic retinopathy treated as the exposure. Population: adults with type 1 or type 2 diabetes from diabetes-specific cohorts, or participants in population-based cohorts providing an explicitly diabetic-retinopathy–specific estimate; Exposure: diabetic retinopathy identified through clinical examination, retinal photography, validated grading systems, medical records, or diagnostic codes; Comparator: participants without diabetic retinopathy or the study-defined reference group for the diabetic-retinopathy–specific estimate; Outcomes: cognitive impairment, mild cognitive impairment, all-cause dementia, Alzheimer’s disease, or vascular dementia assessed using validated cognitive tests, established diagnostic criteria, medical records, or diagnostic codes; and Study design: full-text, English-language cross-sectional, prospective cohort, or retrospective cohort studies. Quantitative synthesis required sufficient event data to calculate an odds ratio or a reported hazard ratio with a corresponding measure of uncertainty. We excluded non-human studies, duplicate reports, case reports, case series, conference abstracts, letters, editorials, commentaries, reviews, meta-analyses, and studies without an eligible exposure, comparator, cognitive outcome, or extractable effect estimate. Studies reporting only nonspecific retinopathy without establishing a diabetic etiology or providing a diabetic-retinopathy–specific estimate were also excluded.

2.2.Information sources and search strategy

A preliminary scoping search of PubMed, Scopus, and Embase was used only to identify candidate keywords and controlled-vocabulary terms. Embase was not part of the formal systematic retrieval set and was therefore not included in the PRISMA record counts. The formal systematic search was conducted in PubMed, Scopus, Web of Science, and the Cochrane Library from database inception through December 2025. The strategies combined controlled-vocabulary and free-text terms describing diabetes mellitus, diabetic retinopathy or diabetic retinal disease, cognitive impairment, cognitive decline, mild cognitive impairment, dementia, Alzheimer’s disease, and vascular dementia. The complete database-specific strategies, including search dates, fields, Boolean operators, filters, and limits, are provided in Supplementary Table S2. Backward and forward citation searching was also performed for eligible reports. Trial registries, theses, dissertations, and preprint servers were not systematically searched.

2.3.Selection process

Two reviewers independently screened the titles and abstracts of all unique records against the predefined eligibility criteria. The same reviewers then independently assessed the full texts of potentially eligible reports. Disagreements at either stage were resolved through discussion and consensus; when consensus could not be reached, a third reviewer adjudicated the decision.

2.4.Data collection

Two reviewers independently extracted data into a standardized shared spreadsheet using Google Sheets. Extracted items included first author, publication year, country, study design, data source and study period, sample size, participant age and sex, diabetes type, definition and severity of diabetic retinopathy, comparator definition, cognitive outcome and ascertainment method, follow-up duration, effect measure, adjusted covariates, and quantitative data required for synthesis, including event counts, odds ratios, hazard ratios, 95% confidence intervals, and p values. After extraction, the reviewers cross-checked the entries and resolved discrepancies by consensus; unresolved disagreements were adjudicated by a third reviewer. Missing or unclear information was recorded as not reported.

When a study reported multiple eligible estimates, we selected the most fully adjusted model that retained the eligible DR exposure, comparator, and cognitive outcome. DR-specific estimates were preferred over composite diabetic-eye-disease estimates, and diabetes-restricted estimates were preferred when the overall cohort was population based. For Lee et al., the time-dependent DMR estimate from Model 3 was selected. For Shang et al., the DR-specific analysis restricted to participants with diabetes was selected rather than the broader DRED estimate. The study-level numerical data and the source of every estimate used in the quantitative syntheses are presented in Table 1. Accordingly, the Hwang and Lee population-based cohorts were retained because they reported explicit DR/DMR estimates; they were not treated as direct DR-versus-no-DR comparisons within exclusively diabetes-only cohorts.

Table 1Study-level numerical data used in the quantitative syntheses of diabetic retinopathy and cognitive outcomes.

Study Outcome Exposure Comparator DR events/total Comparator events/total Effect used (95% CI) Estimate type/source Weight

Crosby-Nwaobi 2013 (25) Positive cognitive screen PDR No/mild DR 6/128 31/252 OR 0.35 (0.14–0.86) Crude OR derived from Table 3 2.2%

Gao 2016 (30) Dementia DR No DR 40/325 92/784 OR 1.06 (0.71–1.57) Crude OR derived from Table 2 11.4%

Kimura 2008 (33) Dementia DR No DR 14/106 32/280 OR 1.18 (0.60–2.31) Crude OR derived from Table 2 4.0%

Murata 2017 (27) MCI or dementia Retinopathy No retinopathy 20/51 58/196 OR 1.54 (0.81–2.91) Crude OR derived from Table 1 4.4%

Xia 2020 (28) Dementia DR No DR 28/146 19/151 OR 1.65 (0.88–3.11) Crude OR derived from Table 2 4.5%

Bruce 2014 (24) Incident dementia Baseline DR No baseline DR 4/30 13/290 OR 3.28 (1.00–10.78) Crude OR derived from Table 3 1.3%

Hendrie 2018 (31) Dementia Retinopathy No retinopathy 302/1,698 611/3,530 OR 1.03 (0.89–1.20) Crude OR derived from Table 3 72.2%

Deal 2019 (26) Incident all-cause dementia Moderate/severe DR No retinopathy — — HR 1.86 (1.36–2.55) Adjusted diabetes-restricted estimate 13.8%

Exalto 2014 (29) Incident dementia Severe diabetic retinal disease Absent — — HR 1.37 (1.22–1.53) Adjusted model, Table 4 16.5%

Lee 2025 (37) Incident dementia Time-dependent DMR Study-defined no-DMR reference group — — HR 2.74 (2.44–3.08) Model 3; original value 2.738 (2.437–3.075) 16.5%

Rodill 2018 (34) Incident dementia Time-updated DR No DR — — HR 1.12 (0.82–1.53) Adjusted Model 3 13.8%

Yu 2020 (36) Incident all-cause dementia Coded DR No coded DR — — HR 1.09 (1.07–1.11) Fully adjusted model 17.0%

Hwang 2021 (32) Incident all-cause dementia Recorded DR Study-defined no-recorded-DR reference group — — HR 1.51 (0.70–3.27) Adjusted population-based estimate 7.0%

Shang 2023 (35) Incident dementia DR among participants with diabetes No DR among participants with diabetes — — HR 1.61 (1.30–2.00) DR-specific diabetes analysis 15.3%

2.5.Outcome measures

2.5.1.Cognitive impairment

Cognitive outcomes included objectively assessed cognitive impairment, mild cognitive impairment (MCI), all-cause dementia, Alzheimer’s disease, and vascular dementia. MCI and dementia were not considered clinically equivalent. The crude dichotomous synthesis used one eligible study-defined cognitive endpoint from each report and was therefore interpreted as a broad exploratory synthesis of cognitive impairment or dementia. The longitudinal synthesis used adjusted estimates for incident dementia. Alzheimer’s disease and vascular dementia were summarized separately because insufficient comparable data were available for subtype-specific meta-analysis. Eligible outcomes were ascertained using validated cognitive instruments, established clinical criteria, medical records, or diagnostic codes. Some studies defined CI based on a clinical score cutoff, such as the clinical dementia rating (CDR), with scores of 0.5 indicating CI without dementia, and scores of 1 or more indicating dementia. Other assessment tools, such as the Mini-Mental State Examination (MMSE) score, the Addenbrooke’s Cognitive Examination-Revised (ACE-R), or the Montreal Cognitive Assessment (MoCA) score, were used as well (24–28). Other studies used ICD diagnosis codes to define CI/dementia, the DSM-IV criteria for dementia, or the NINCDS-ADRDA criteria for Alzheimer’s disease, or the NINDS-AIREN criteria for VaD (29–36).

2.5.2.Diabetic retinopathy

DR was defined by grades detected on ophthalmoscopy or retinal photography (24, 30, 33), with the ETDRS criteria (25), modified Airlie House classification (26), or by specific diagnostic codes provided by the ICD-9 (29, 31, 32, 34, 36, 37).

2.6.Risk of bias

Because all eligible studies were observational and no randomized controlled trials were included, the Critical Appraisal Skills Programme Randomized Controlled Trial Checklist was not applicable. Cohort and cross-sectional studies were assessed using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (38). To ensure uniform application, two independent reviewers first used the tools on a random sample of studies and refined the assessment criteria. Following this calibration, the same two reviewers independently assessed all studies. Any disagreements were resolved through arbitration by a third reviewer.

2.7.Data synthesis and statistical analysis

Crude event-count data and adjusted time-to-event estimates were synthesized separately because they represent different estimands and should not be interpreted as equivalent. For the crude dichotomous synthesis, study-level 2 × 2 data were converted to log odds ratios and corresponding sampling variances using the escalc function in the metafor package. Adjusted hazard ratios and their confidence limits were log-transformed, and standard errors were derived as [ln(upper confidence limit) − ln(lower confidence limit)]/(2 × 1.96). Both syntheses were fitted using inverse-variance random-effects models with restricted maximum-likelihood estimation of τ2 and were subsequently back-transformed for presentation. The Hartung–Knapp adjustment was not applied. Heterogeneity was described using I2, τ2, the χ2 test, and 95% prediction intervals; because the χ2 test has limited power with few studies, interpretation emphasized I2, τ2, prediction intervals, and clinical and methodological differences. Statistical significance was defined as a two-sided p<0.05.

Adjusted cross-sectional estimates were not sufficiently and consistently reported on comparable scales; therefore, the dichotomous synthesis was limited to crude event data and interpreted as exploratory. ORs and HRs were not combined in a single meta-analysis. A leave-one-out sensitivity analysis was performed for the longitudinal synthesis by refitting the restricted maximum-likelihood model after omitting each study in turn. Because only seven studies contributed to each synthesis and potential subgroups frequently contained one or two studies, post hoc subgroup analysis and meta-regression were not considered sufficiently stable to support reliable inference. Funnel-plot asymmetry was not assessed because each synthesis contained fewer than 10 studies.

3.Results

3.1.Study selection

The database searches identified 288 records. After removal of 21 duplicates, 267 unique records underwent title and abstract screening, of which 214 were excluded. Fifty-three full-text reports were assessed for eligibility. Thirty-nine reports were excluded: 17 had no eligible cognitive outcome, 12 had no eligible DR exposure, two were conference abstracts, seven were reviews, and one reported nonspecific retinopathy without an eligible DR-specific estimate. Fourteen studies were included in the systematic review and quantitative syntheses (24–37). The flow diagram for study selection is shown in Figure 1.

3.2.Characteristics of the included studies

The 14 included studies were published between 2008 and 2025 and comprised five cross-sectional and nine longitudinal observational studies, with reported study populations totaling 2,871,847 participants. Follow-up in the longitudinal studies ranged from 4.1 to 15.6 years. Study-level age, sex, population, exposure, and follow-up characteristics are presented in Table 2. Hwang et al. and Lee et al. were population-based cohorts with study-defined DR/DMR estimates rather than direct comparisons within exclusively diabetic cohorts.

Table 2Baseline characteristics of the included studies.

Study Study design Sample size Country Data source Study period Age Female, n (%) Follow up Population and eligible analysis Exposure Adjustments Conclusion

Bruce 2014 (24) Longitudinal observational study 335 Australia Fremantle Diabetes Study (FDS) Phase I: 1993–1996; Phase II (Follow-up): 2008–2011 57.5 (9.2) 167 (49.3) 14.7 y Diabetes-only cohort: adults with type 2 diabetes DR Age, sex, education, marital status, BMI, stroke/TIA history, diabetes duration, fasting glucose, insulin therapy, and smoking. Insulin therapy, diabetic retinopathy, and smoking in mid-life are independent predictors of cognitive impairment and dementia in patients with Type 2 diabetes.

Crosby-Nwaobi 2013 (25) Cross-sectional study 380 United Kingdom South East London Diabetic Retinopathy Study (SEL-DRS)/Diabetes Eye Complication Service (DECS) NA 64.8 (10.8) 166 (43.8) NA Diabetes-only cohort: adults with type 2 diabetes; PDR compared with no/mild DR DR Age, sex, ethnicity, education, BCVA, diabetes duration, SES, nephropathy, BMI, blood pressure, HbA1c, lipids, alcohol, depression, and emotional distress. There is an inverse relationship between retinopathy severity and cognitive impairment; patients with proliferative retinopathy (PDR) had better cognitive performance than those with no/mild retinopathy.

Deal 2019 (26) Population-based prospective cohort study 12,482 United States Atherosclerosis Risk in Communities (ARIC) Study Baseline (Visit 3): 1993–1995; Follow-up through 2011–2013 50–73 y 6,929 (56) 15.6 y Population-based cohort; diabetes-restricted moderate/severe retinopathy estimate used DR Age, sex, race, education, BMI, smoking, alcohol, hypertension, diabetes, coronary heart disease, stroke, APOE genotype. Moderate/severe retinopathy and generalized arteriolar narrowing in mid-life are associated with an increased risk of incident dementia. Retinopathy was associated with cerebrovascular-related dementia etiology but not Alzheimer’s-related etiology.

Exalto 2014 (29) Longitudinal cohort study 29,961 United States Kaiser Permanente Northern California (KPNC) Diabetes Registry Baseline: 1996–1998; Follow-up: 1998–2008 70.6 (6.8) 13,787 (46) 10 y Diabetes-only cohort: adults aged ≥60 years with type 2 diabetes DR Age, gender, race/ethnicity, education, medical utilization, diabetes duration, HbA1c, pharmacotherapy, hyperglycemia/hypoglycemia events, BMI, smoking, and vascular comorbidities (hypertension, lipids, CVD, etc.). Severe diabetic retinal disease is associated with an increased risk of incident dementia in patients with Type 2 diabetes, suggesting a link between microvascular disease and dementia.

Gao 2016 (30) Cross-sectional study 1,109 China Community-dwelling residents in Tianjin March 2010 – April 2011 72.4 (3.5) y NA NA Diabetes-only cohort: Han Chinese adults aged ≥65 years with type 2 diabetes DR Age, gender, education, income, smoking, alcohol, hypertension, hyperlipidemia, CVD history, BMI, HbA1c, diabetes duration, visual impairment. Conclusion: While Type 2 Diabetes was associated with a high prevalence of Mild Cognitive Impairment (MCI), retinopathy itself was not found to be a statistically significant factor distinguishing between normal cognition, MCI, and dementia groups in the multivariate analysis

Hendrie 2018 (31) Longitudinal cohort study 5228 United States Eskenazi Health Services electronic medical records January 1, 1992 – December 31, 2014 76.6 (4.9) y NA 13.8 y Diabetes-only cohort: 5,228 adults with diabetes; Table 3 provided extractable retinopathy–dementia data DR Not applicable to the pooled crude OR. Among 5,228 participants with diabetes, Table 3 reported dementia in 302 of 1,698 participants with retinopathy and 611 of 3,530 without retinopathy. These unadjusted counts were used in the crude dichotomous synthesis.

Kimura 2008 (33) Cross-sectional study 386 Japan Kariya Toyota General Hospital September – November 2003 66.1 (8.2) 160 (41.5) NA Diabetes-only cohort: adults aged >50 years with diabetes DR Adjusted for age, sex, BMI, diabetes duration The prevalence of retinopathy was 30.4% in demented subjects vs 27.1% in non-demented subjects; this difference was not statistically significant

Lee 2025 (37) Longitudinal retrospective cohort study 784,205 South Korea Korean National Health Insurance Screening Cohort (NHIS-HEALS) 2002–2017 more than 40 years 423,836 (54) 10.9 ± 2.7 y Population-based cohort with explicitly coded DMR exposure and a study-defined reference group; not an exclusively diabetes-only DR-versus-no-DR comparison DR Age, sex, income, visual acuity, systemic comorbidities (HTN, dyslipidemia, stroke, depression, etc.), behavioral factors (smoking, alcohol, BMI), diabetes severity Patients with DMR had a significantly higher risk of developing dementia (HR 2.74). The risk was higher in those with severe DMR and notably higher in younger patients.

Murata 2017 (27) Cross-sectional study 281 Japan Japan Diabetes Clinical Data Management Study (JDDM) NA 73.4 (6.1) y 137 (48.8) NA Diabetes-only cohort: adults aged ≥65 years with type 2 diabetes; retinopathy information available for 247 participants DR Multiple regression for MMSE score included NEFA, eGFR, insulin, age, and schooling Prevalence of retinopathy was not significantly different between dementia (33.3%), MCI (23.9%), and normal cognition (18.4%) groups in this cross-sectional study

Rodill 2018 (34) Longitudinal cohort study 3742 United States Kaiser Permanente Northern California (KPNC) 1996–2015 56.1 (8.5) 1,772 (47.4) 6.2 y Diabetes-only cohort: adults aged ≥50 years with type 1 diabetes DR Age, sex, race, HbA1c, acute glycemic events, neuropathy, nephropathy, ESRD, CVD, stroke In this cohort of older patients with Type 1 diabetes, diabetic retinopathy was not significantly associated with an increased risk of incident dementia (HR 1.12, 95% CI 0.82–1.54)

Xia 2020 (28) Cross-sectional study 297 China Tongji Hospital, Tongji Medical College April 2018 – May 2019 56.8 (6.9) 105 (35.4) NA Diabetes-only cohort: adults aged 45–74 years with type 2 diabetes DR Age, education, diabetes duration, BMI, HbA1c, lipids, and blood pressure. Diabetic retinopathy was identified as a significant risk factor for dementia (OR 2.197, P = 0.040) in this cross-sectional analysis, though it was not related to Mild Cognitive Impairment (MCI)

Yu 2020 (36) Retrospective population-based cohort study 1,917,702 South Korea National Health Insurance Service (NHIS) 2009–2015 56.8 (6.9) 788175 (41.1) 5.1 y Diabetes-only cohort: adults aged ≥40 years with diabetes DR Age, sex, smoking, alcohol, exercise, income, glucose, diabetes duration, BMI, dyslipidemia, hypertension, CKD, stroke, IHD, depression, meds Diabetic retinopathy was significantly associated with the development of all-cause dementia (HR 1.09), Alzheimer’s Disease (HR 1.10), and Vascular Dementia (HR 1.08)

Hwang 2021 (32) Prospective cohort study 3,375 United States Cardiovascular Health Study (CHS) and Medicare Part B claims Baseline 1989-93; Cognition Study 1998-99 74.8 (9.5) NA 4.1 y Population-based cohort with explicitly identified DR exposure and a study-defined reference group; not an exclusively diabetes-only DR-versus-no-DR comparison DR Demographics, cardiovascular risk factors, APOE genotype, and other ophthalmic conditions Diabetic retinopathy was significantly associated with an increased risk of Vascular Dementia/mixed dementia (HR 2.63), but not Alzheimer’s Disease or all-cause dementia in this cohort

Shang 2023 (35) Prospective cohort study 112364 United Kingdom UK Biobank Baseline 2006–2010; Follow-up to 2021 62.4 (7.3) 60676 (54) 11 y Population-based cohort; DR-specific analysis restricted to participants with diabetes used DR Age, gender, education, income, smoking, alcohol, physical activity, BMI, cholesterol, glucose, diet Diabetes-related eye disease (DRED) was independently associated with an increased risk of incident dementia (HR 1.61, 95% CI 1.30–2.00) among patients with diabetes

DR, diabetic retinopathy; DMR, diabetic retinopathy/diabetic retinal disease; PDR, proliferative diabetic retinopathy; MCI, mild cognitive impairment; OR, odds ratio; HR, hazard ratio; CI, confidence interval; NR, not reported. The reported overall cohort size may differ from the analytical denominator contributing to a particular outcome. The crude OR synthesis used event counts extracted from the original reports, whereas the longitudinal synthesis used study-reported adjusted HRs. Population-based studies were retained only when an explicitly DR-related or diabetes-restricted estimate relevant to the review question was available. Random-effects weights are those shown in the corresponding forest plots.

3.3.Risk of bias

Using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies, 12 studies were rated as good quality and two as fair quality; no study was rated poor (Supplementary Table S1; Supplementary Figure S1).

3.4.Results of synthesis

3.4.1.Quantitative syntheses of diabetic retinopathy and cognitive outcomes

Seven studies, Bruce, Crosby-Nwaobi, Gao, Hendrie, Kimura, Murata, and Xia, contributed crude dichotomous event data. The pooled analysis did not demonstrate a statistically significant association between DR and the study-defined cognitive impairment or dementia outcomes (OR 1.07, 95% CI 0.94–1.23; p=0.310; I2=52.3%; τ2=0.0005; 95% prediction interval 0.90–1.28; Figure 2; Table 1). Seven studies, Deal, Exalto, Lee, Rodill, Yu, Hwang, and Shang, contributed adjusted longitudinal hazard estimates for incident dementia. Each study contributed an incident dementia endpoint as defined in its report; the study-specific endpoint and ascertainment are shown in Tables 2 and 1. DR was associated with a higher incidence of dementia (HR 1.55, 95% CI 1.19–2.02; p=0.001; I2=97.8%; τ2=0.1073; 95% prediction interval 0.65–3.68; Figure 3; Table 1). The prediction interval crossed the null value, indicating that the magnitude and statistical significance of the association may differ substantially across future settings. Leave-one-out pooled estimates ranged from HR 1.36 (95% CI 1.13–1.63) after omitting Lee et al. to HR 1.66 (95% CI 1.26–2.20) after omitting Yu et al.; no single-study omission changed the direction or statistical significance of the association (Supplementary Table S3).

3.4.2.Association between diabetic retinopathy and dementia subtypes

Evidence concerning dementia subtypes was inconsistent. Hwang et al. reported that DR was associated with vascular or mixed dementia (HR 2.63, p=0.029) but not Alzheimer’s disease (HR 0.46, p=0.439); however, these estimates were based on relatively few DR-exposed participants and events (32). Deal et al. similarly reported an association between moderate/severe DR and cerebrovascular-related dementia or MCI (HR 2.29, p=0.008), whereas the estimate for Alzheimer’s-related dementia or MCI was not statistically significant (HR 1.33, p=0.452) (26). This pattern is compatible with a small-vessel-disease hypothesis but does not establish an underlying mechanism.

Conversely, Lee et al. reported associations with both Alzheimer’s disease (HR 2.89, 95% CI 2.50–3.34) and vascular dementia (HR 2.37, 95% CI 1.40–4.02) (37). Yu et al. also reported associations with Alzheimer’s disease (HR 1.10, 95% CI 1.07–1.12) and vascular dementia (HR 1.08, 95% CI 1.03–1.14) (36). Because definitions, populations, and analytical models differed across studies, these findings were not pooled and do not establish a consistent subtype-specific association.

4.Discussion

The present meta-analysis found no statistically significant association in the crude dichotomous synthesis, whereas adjusted longitudinal evidence indicated an association between diabetic retinopathy and a higher incidence of dementia (HR 1.55). The longitudinal synthesis included incident dementia outcomes rather than general cognitive impairment. The Hwang and Lee estimates arose from population-based cohorts and should not be interpreted as direct diabetes-only DR-versus-no-DR comparisons. This temporal distinction suggests that DR may be a prognostic marker for future cognitive deterioration rather than merely coexisting with established impairment, potentially reflecting shared pathophysiology that precedes clinically detectable cognitive deficits. In a systematic review, Dove et al. found that, after a mean follow-up duration of 9.2 years, 29.6% of diabetic patients who were cognitively intact at baseline developed CI without dementia (CIND), which is a category for those who show signs of cognitive decline without fitting the criteria for dementia. In the CIND-at-baseline cohort, 22.1% were diagnosed with dementia after a mean follow-up duration of 7.7 years. They noted that progression to dementia was associated with worse glycemic control (HbA1c >7.5%; HR 2.87, 95% CI: 1.20; 6.85) (4).

In a prior systematic review, Chan et al. identified a statistically significant association between DR and any type of cognitive decline from five studies with longitudinal data (Relative risk [RR] 1.31). From cross-sectional data, 18 studies found that the odds of any CI (OR 1.48) and early-stage CI (OR 1.59) were significantly higher among patients with DR compared to those without DR; however, significant heterogeneity limited the reliability of these estimates (19). Similarly, Cheng et al. pooled data from ten studies and detected a statistically meaningful increase in the odds of CI in patients with DR (OR 2.24, 95% CI: 1.89; 2.66). Subgrouping of studies based on adjustment for confounders revealed similar odds of CI with DR (OR 2.12, 95% CI: 1.55; 2.88). In two studies, where patients were followed up for more than ten years, the odds were significantly higher (OR 2.92) with no detectable heterogeneity. Nevertheless, these estimates were pooled from a small number of studies encompassing a relatively small number of patients, so cautious interpretation is warranted (18). Based on data from four studies, Feng et al. reported a pooled relative risk of AD of 1.21 among patients with DR (95% CI: 1.04; 1.41). Conversely, the association between DR and the risk of VaD did not reach statistical significance in the pooled analysis (RR 1.44, 95% CI: 0.93; 2.23) (20). Differences between those estimates and the present findings may reflect our exclusion of nonspecific retinopathy, verification of DR-specific or diabetes-restricted estimates in population-based cohorts, separation of crude dichotomous ORs from adjusted longitudinal HRs, and inclusion of the 2025 Lee cohort. ORs, RRs, and HRs are not mathematically equivalent and were therefore compared only contextually; in particular, an OR may overstate an RR when the outcome is common.

The observed association may reflect shared retinal and cerebral microvascular pathology. Previous studies have associated DR with cerebral small-vessel-disease markers, structural brain changes, and subsequent cognitive decline (39–42). Hyperglycemia-related endothelial dysfunction, oxidative stress, inflammation, vascular endothelial growth factor, and advanced glycation end products have been proposed as processes affecting both retinal and cerebral neurovascular tissues (43, 44). Neurodegenerative pathways involving amyloid and tau have also been proposed (10, 11, 45–48); however, these biomarkers were not evaluated in the present meta-analysis, and these mechanisms should be regarded as hypotheses rather than findings of this review. The available evidence does not establish causality or demonstrate that routine cognitive screening based solely on DR status improves clinical outcomes.

Our findings regarding dementia subtypes clarified a relationship between microvascular and neurodegenerative processes. Several large population-based studies identified strong associations between DR and VaD or cerebrovascular-related CI. These observations align with the retinal/cerebral small vessel disease hypothesis, which posits that DR represents a visible marker of systemic microangiopathy affecting end-organs, including the brain (41). However, conflicting evidence from nationwide Korean cohort studies challenges this dichotomous perspective. Lee et al. and Yu et al. reported significant associations between DR and both AD and VaD, with comparable effect sizes across subtypes (36, 37).

Gupta et al. hypothesized that the shared embryological origin of cerebral and retinal blood barriers may make them both vulnerable to similar deterioration in diabetes (42). Another hypothesis reported in the work of Arden et al. indicates that hyperglycemia in diabetes leads to endothelial dysfunction, causing hypoxia and ischemia by means of vascular endothelial growth factor (VEGF), advanced glycation end products (AGEs), oxidative stress, and inflammatory processes, which occur both in the cerebral and retinal vessels (43). Gorska-Ciebiada et al. inferred a notable elevation in serum AGEs in diabetic patients with MCI compared to controls. Receptors for AGEs (RAGE) bind AGEs and inflate the extent of oxidative stress and inflammation by activation of nuclear factor kappa B (NF-kB) in neurovascular tissues in the retina and the brain (44).

There are various hypotheses regarding the interaction between diabetes and AD. For instance, the hyperinsulinemic state in early type 2 DM has been implicated in the deposition of amyloid-beta in neural tissues due to the consumption of insulin-degrading enzyme (IDE) by high insulin levels. IDE is responsible for the degradation and clearance of amyloid-beta, the accumulation of which leads to the formation of senile plaques in the brain, causing AD (45–47). Moreover, Zhu et al. posited another mechanism related to the other histopathological aspect of AD, the neurofibrillary tangles (NFTs), which consist of altered microtubule-related tau protein (48). Their hypothesis indicates that glycogen synthase kinase 3β (GSK3β)-mediated hyperphosphorylation of tau protein, in patients with DR, contributes to synaptic degeneration in retinal ganglion cells (RGCs) prior to microvascular changes, similar to cerebral changes in AD (11). This shared neurotoxic mechanism seems plausible considering that retinal neurodegenerative changes occur much earlier than microvascular changes in the diabetic eye (10).

From a research perspective, the heterogeneity in subtype-specific associations emphasized the need for standardized neuropathological confirmation in epidemiologic studies. Future investigations should employ multimodal biomarker approaches, including amyloid and tau imaging, vascular MRI markers, and retinal imaging with quantitative microvascular assessment, to disentangle the relative contributions of vascular and neurodegenerative pathology. Several limitations should be considered. First, all included studies were observational; residual and unmeasured confounding cannot be excluded, and the covariates included in adjusted models varied substantially. Second, considerable clinical and methodological heterogeneity existed in population composition, DR definition and severity, dementia definition and ascertainment, follow-up duration, and adjustment strategy. In particular, Hwang et al. and Lee et al. were population-based cohorts using study-defined reference groups, which limits direct comparability with diabetes-only cohorts. Only one included study specifically evaluated type 1 diabetes, and the evidence therefore predominantly reflects type 2 diabetes. The inconsistent severity categories prevented a reliable dose–response synthesis; although Lee et al. reported associations for both moderate and severe DMR, the available studies did not demonstrate a consistent severity gradient. Third, only seven studies contributed to each quantitative synthesis. Although the leave-one-out analysis did not identify a single influential study, reliable subgroup analysis and meta-regression were not performed, and the extreme longitudinal heterogeneity and prediction interval crossing the null limit confidence in the pooled estimate. Fourth, formal funnel-plot asymmetry testing was underpowered, and publication bias could not be adequately assessed. A formal GRADE certainty assessment was not undertaken, which should also be considered when interpreting the strength of the evidence. Fifth, administrative coding and medical-record ascertainment in several studies may have introduced both exposure and outcome misclassification. Finally, restriction to English-language full-text publications and the absence of a systematic grey-literature search may have introduced language and publication biases.

5.Conclusion

In conclusion, the crude dichotomous synthesis did not demonstrate a statistically significant association between DR and cognitive impairment or dementia, whereas adjusted longitudinal evidence suggested that DR was associated with a higher incidence of dementia. However, the observational study designs, residual confounding, extreme longitudinal heterogeneity, a prediction interval crossing the null, and possible exposure and outcome misclassification limit confidence and require cautious interpretation. Evidence concerning Alzheimer’s disease and vascular dementia remains inconsistent. These findings do not establish causality or the benefit of routine cognitive screening based solely on DR status.

Funding Statement

The author(s) declared that financial support was not received for this work and/or its publication.

Footnotes

Edited by: Teodoro Cordova, University of Guanajuato, Mexico

Reviewed by: Jose Alejandro Valdevila Figueira, Universidad Tecnológica Ecotec, Ecuador

Vignesh D., ESIC Medical College and Postgraduate Institute of Medical Science and Research, India

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Author contributions

OS: Data curation, Writing – original draft, Conceptualization, Methodology, Formal analysis, Investigation, Writing – review & editing. WA: Data curation, Conceptualization, Writing – review & editing, Methodology, Formal analysis, Investigation, Writing – original draft. TA: Data curation, Validation, Writing – review & editing, Investigation. AbMA: Data curation, Validation, Writing – review & editing, Investigation. MSA: Data curation, Writing – review & editing, Investigation, Validation. AAMA: Validation, Writing – review & editing, Investigation, Data curation. AIA: Investigation, Data curation, Validation, Writing – review & editing. BA: Data curation, Investigation, Writing – review & editing. RA: Writing – review & editing, Investigation, Data curation. AhMA: Validation, Writing – review & editing. MEA: Writing – review & editing, Validation. SA: Writing – review & editing, Validation. IB: Conceptualization, Writing – review & editing, Supervision, Validation, Methodology.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fendo.2026.1881602/full#supplementary-material

References

1.
Ogurtsova K, da Rocha Fernandes JD, Huang Y, Linnenkamp U, Guariguata L, Cho NH, et al.
IDF Diabetes Atlas: global estimates for the prevalence of diabetes for 2015 and 2040. Diabetes Res Clin Pract. (2017) 128:40–50. doi:  10.1016/j.diabres.2017.03.024

2.
Raghuveer P, Shamsudeen M, Sridhar S, Singla R, Gupta V, Mittal N, et al.
Prevalence and factors associated with cognitive impairment among persons with type 2 diabetes mellitus: a cross-sectional study in medical college hospitals of South India. Clin Epidemiol Global Health. (2025) 31:101887. doi:  10.1016/j.cegh.2024.10188742574925

3.
Ehtewish H, Arredouani A, El‐Agnaf OMA.
Diagnostic, prognostic, and mechanistic biomarkers of diabetes mellitus-associated cognitive decline. Int J Mol Sci. (2022) 23:6144. doi:  10.3390/ijms23116144

4.
Dove A, Shang Y, Xu W, Grande G, Laukka EJ, Fratiglioni L, et al.
The impact of diabetes on cognitive impairment and its progression to dementia. Alzheimer S Dementia. (2021) 17:1769–78. doi:  10.1002/alz.12482

5.
Samaras K, Sachdev PS.
Diabetes and the elderly brain: sweet memories? Ther Adv Endocrinol Metab. (2012) 3:189–96. doi:  10.1177/2042018812469645

6.
Fiore V, De Rosa A, Falasca P, Marci M, Guastamacchia E, Licchelli B, et al.
Focus on the correlations between Alzheimer’s disease and type 2 diabetes. Endocr Metab Immune Disord - Drug Targets. (2019) 19:571–9. doi:  10.2174/1871530319666190311141855

7.
Wang G, Li W.
Diabetes as a risk factor for abnormal cognition development. J Alzheimer S Dis Rep. (2020) 4:237–42. doi:  10.3233/adr-200181

8.
Maxwell SS, Thomas WG.
Neurodegeneration in the pathogenesis of diabetic retinopathy: molecular mechanisms and therapeutic implications. Curr Med Chem. (2013) 20:3241–50. doi:  10.2174/09298673113209990027

9.
Carmichael J, Fadavi H, Tavakoli M.
Neurodegeneration of the cornea and retina in patients with type 1 diabetes without clinical evidence of diabetic retinopathy. Front Endocrinol. (2022) 13. doi:  10.3389/fendo.2022.790255

10.
Sohn EH, van Dijk HW, Jiao C, Kok PHB, Jeong W, Demirkaya N, et al.
Retinal neurodegeneration may precede microvascular changes characteristic of diabetic retinopathy in diabetes mellitus. Proc Natl Acad Sci. (2016) 113:E2655–64. doi:  10.1073/pnas.1522014113

11.
Zhu H, Zhang W, Zhao Y, Shu X, Wang W, Wang D, et al.
GSK3β-mediated tau hyperphosphorylation triggers diabetic retinal neurodegeneration by disrupting synaptic and mitochondrial functions. Mol Neurodegener. (2018) 13:62. doi:  10.1186/s13024-018-0295-z

12.
del Zoppo GJ.
The neurovascular unit in the setting of stroke. J Intern Med. (2010) 267:156–71. doi:  10.1111/j.1365-2796.2009.02199.x

13.
Zlokovic BV.
The blood-brain barrier in health and chronic neurodegenerative disorders. Neuron. (2008) 57:178–201. doi:  10.1016/j.neuron.2008.01.003

14.
Sinclair AJ, Vellas B.
Diabetes mellitus and cognitive decline – prevention should not be delayed! J Prev Alzheimer S Dis. (2018) 5:95–7. doi:  10.14283/jpad.2018.9

15.
Wrighten SA, Piroli GG, Grillo CA, Reagan LP.
A look inside the diabetic brain: contributors to diabetes-induced brain aging. Biochim Et Biophys Acta (Bba) - Mol Basis Dis. (2009) 1792:444–53. doi:  10.1016/j.bbadis.2008.10.013

16.
Zhou X, Zhang J, Chen Y, Ma T, Wang Y, Wang J, et al.
Aggravated cognitive and brain functional impairment in mild cognitive impairment patients with type 2 diabetes: a resting-state functional MRI study. J Alzheimer’s Dis. (2014) 41:925–35. doi:  10.3233/jad-132354

17.
Cui Y, Jiao Y, Chen YC, Wang K, Gao B, Wen S, et al.
Altered spontaneous brain activity in type 2 diabetes: a resting-state functional MRI study. Diabetes. (2014) 63:749–60. doi:  10.2337/db13-0519

18.
Cheng D, Zhao X, Yang S, Wang G, Ning G.
Association between diabetic retinopathy and cognitive impairment: a systematic review and meta-analysis. Front Aging Neurosci. (2021) 13. doi:  10.3389/fnagi.2021.692911

19.
Chan RNF, Tang Z, Chan VTT, Chan RNC, Cheng ETW, Ng NCY, et al.
The cross-sectional and longitudinal relationship of diabetic retinopathy to cognitive impairment: a systematic review and meta-analysis. Eye. (2023) 37:220–7. doi:  10.1038/s41433-022-02033-2

20.
Feng J, Huang C, Liang L, Li C, Wang X, Ma J, et al.
The association between eye disease and incidence of dementia: systematic review and meta-analysis. J Am Med Directors Assoc. (2023) 24:1363–1373.e6. doi:  10.1016/j.jamda.2023.06.025

21.
Kuzma E, Littlejohns TJ, Khawaja AP, Llewellyn DJ, Ukoumunne OC, Thiem U, et al.
Visual impairment, eye diseases, and dementia risk: a systematic review and meta-analysis. J Alzheimer’s Dis. (2021) 83:1073–87. doi:  10.3233/JAD-210250

22.
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al.
The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Bmj. (2021) 372:n71. doi:  10.1136/bmj.n71

23.
Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, et al. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.5, Updated August 2024 (2024). Available online at: https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current (Accessed October 3, 2026).

24.
Bruce DG, Davis WA, Starkstein SE, Davis TME.
Mid-life predictors of cognitive impairment and dementia in type 2 diabetes mellitus: the Fremantle Diabetes Study. J Alzheimer's Dis. (2014) 42:S63–70. doi:  10.3233/jad-132654

25.
Crosby-Nwaobi RR, Sivaprasad S, Amiel S, Forbes A.
The relationship between diabetic retinopathy and cognitive impairment. Diabetes Care. (2013) 36:3177–86. doi:  10.2337/dc12-2141

26.
Deal JA, Sharrett AR, Albert M, Bandeen-Roche K, Burgard S, Thomas SD, et al.
Retinal signs and risk of incident dementia in the Atherosclerosis Risk in Communities study. Alzheimer's Dementia. (2019) 15:477–86. doi:  10.1016/j.jalz.2018.10.002

27.
Murata Y, Kadoya Y, Yamada S, Sanke T.
Cognitive impairment in elderly patients with type 2 diabetes mellitus: prevalence and related clinical factors. Diabetol Int. (2017) 8:193–8. doi:  10.1007/s13340-016-0292-9

28.
Xia SS, Xia WL, Huang JJ, Zou HJ, Tao J, Yang Y.
The factors contributing to cognitive dysfunction in type 2 diabetic patients. Ann Trans Med. (2020) 8:104. doi:  10.21037/atm.2019.12.113

29.
Exalto LG, Biessels GJ, Karter AJ, Huang ES, Quesenberry CP Jr, Whitmer RA.
Severe diabetic retinal disease and dementia risk in type 2 diabetes. J Alzheimer's Dis. (2014) 42:S109–17. doi:  10.3233/jad-132570

30.
Gao Y, Xiao Y, Miao R, Zhao J, Cui M, Huang G, et al.
The prevalence of mild cognitive impairment with type 2 diabetes mellitus among elderly people in China: a cross-sectional study. Arch Gerontology Geriatrics. (2016) 62:138–42. doi:  10.1016/j.archger.2015.09.003

31.
Hendrie HC, Zheng M, Lane KA, Ambuehl R, Purnell C, Li S, et al.
Changes of glucose levels precede dementia in African‐Americans with diabetes but not in Caucasians. Alzheimer's Dementia. (2018) 14:1572–9. doi:  10.1016/j.jalz.2018.03.008

32.
Hwang PH, Longstreth WT Jr, Thielke SM, Francis CE, Carone M, Kuller LH, et al.
Ophthalmic conditions associated with dementia risk: the Cardiovascular Health Study. Alzheimer's Dementia. (2021) 17:1442–51. doi:  10.1002/alz.12313

33.
Kimura R, Tomiyasu H, Takeuchi T, Shimizu M, Hayashi Y, Okayama N, et al.
Prevalence of Alzheimer's disease with diabetes in the Japanese population. Psychogeriatrics. (2008) 8:73–8. doi:  10.1111/j.1479-8301.2008.00229.x40046247

34.
Rodill LG, Exalto LG, Gilsanz P, Biessels GJ, Quesenberry CP Jr, Whitmer RA.
Diabetic retinopathy and dementia in type 1 diabetes. Alzheimer Dis Associated Disord. (2018) 32:125–30. doi:  10.1097/wad.0000000000000230

35.
Shang X, Zhu Z, Huang Y, Zhang X, Wang W, Shi D, et al.
Associations of ophthalmic and systemic conditions with incident dementia in the UK Biobank. Br J Ophthalmol. (2023) 107:275–82. doi:  10.1136/bjophthalmol-2021-319508

36.
Yu JH, Han K, Park S, Cho H, Lee DY, Kim JW, et al.
Incidence and risk factors for dementia in type 2 diabetes mellitus: a nationwide population-based study in Korea. Diabetes Metab J. (2020) 44:113. doi:  10.4093/dmj.2018.0216

37.
Lee C, Hwang S, Kim M, Bae Y, Chung TY, Jang H, et al.
The influence of diabetic retinopathy on the risk of dementia: a nationwide cohort study. Sci Rep. (2025) 15:38024. doi:  10.1038/s41598-025-21868-9

38.
Health, NIo . Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (2014). Available online at: https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools (Accessed October 3, 2026).

39.
Sanahuja J, Alonso N, Diez J, Ortega E, Rubinat E, Traveset A, et al.
Increased burden of cerebral small vessel disease in patients with type 2 diabetes and retinopathy. Diabetes Care. (2016) 39:1614–20. doi:  10.2337/dc15-2671

40.
Hugenschmidt CE, Lovato JF, Ambrosius WT, Bryan RN, Gerstein HC, Horowitz KR, et al.
The cross-sectional and longitudinal associations of diabetic retinopathy with cognitive function and brain MRI findings: the Action to Control Cardiovascular Risk in Diabetes (ACCORD) Trial. Diabetes Care. (2014) 37:3244–52. doi:  10.2337/dc14-0502

41.
Umemura T, Kawamura T.
Retinopathy: a sign of cerebral small vessel disease in diabetes? J Diabetes Investig. (2017) 8:428–30. doi:  10.1111/jdi.12602

42.
Gupta P, Gan ATL, Man REK, Fenwick EK, Sabanayagam C, Mitchell P, et al.
Association between diabetic retinopathy and incident cognitive impairment. Br J Ophthalmol. (2019) 103:1605. doi:  10.1136/bjophthalmol-2018-312807

43.
Arden GB, Sivaprasad S.
Hypoxia and oxidative stress in the causation of diabetic retinopathy. Curr Diabetes Rev. (2011) 7:291–304. doi:  10.2174/157339911797415620

44.
Gorska-Ciebiada M, Saryusz-Wolska M, Ciebiada M, Loba J.
Mild cognitive impairment and depressive symptoms in elderly patients with diabetes: prevalence, risk factors, and comorbidity. J Diabetes Res. (2014) 2014:179648. doi:  10.1155/2014/179648

45.
Sinha S, Lieberburg I.
Cellular mechanisms of beta-amyloid production and secretion. Proc Natl Acad Sci USA. (1999) 96:11049–53. doi:  10.1073/pnas.96.20.11049

46.
Edbauer D, Willem M, Lammich S, Steiner H, Haass C.
Insulin-degrading enzyme rapidly removes the beta-amyloid precursor protein intracellular domain (AICD). J Biol Chem. (2002) 277:13389–93. doi:  10.1074/jbc.m111571200

47.
Craft S, Peskind E, Schwartz MW, Schellenberg GD, Raskind M, Porte D Jr.
Cerebrospinal fluid and plasma insulin levels in Alzheimer's disease: relationship to severity of dementia and apolipoprotein E genotype. Neurology. (1998) 50:164–8. doi:  10.1212/wnl.50.1.164

48.
Lee VM, Balin BJ, Otvos L Jr, Trojanowski JQ.
A68: a major subunit of paired helical filaments and derivatized forms of normal Tau. Science. (1991) 251:675–8. doi:  10.1126/science.1899488

Associated Data

Supplementary Materials

Data Availability Statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.