Alzheimer’s disease (AD) is the predominant cause of dementia, accounting for approximately 60-80% of cases. Its primary pathological hallmarks are β-amyloid (Aβ) plaques and neurofibrillary tangles composed of hyperphosphorylated tau, which can impair synaptic plasticity and lead to neuronal death (Begcevic et al., 2018). Clinical AD is preceded by a prolonged asymptomatic phase, termed preclinical AD, characterized by the accumulation of brain pathology, which may begin 10-20 years before cognitive symptoms emerge (Bateman et al., 2012; Ameri et al., 2020). The progression of AD includes mild cognitive impairment (MCI), characterized by cognitive decline exceeding age-appropriate and education-appropriate norms but not yet significantly affecting daily activities (Albert et al., 2011). Considering the irreversible nature of AD, early detection of MCI is crucial. Current diagnostic methods for probable AD primarily rely on neurocognitive tests, brain imaging, and cerebrospinal fluid (CSF) analysis (Veitch et al., 2022). However, these procedures are often expensive and invasive for patients. Furthermore, despite these diagnostic tools, AD diagnoses remain inaccurate in 10-15% of cases owing to limitations in sensitivity and specificity (Weller and Budson, 2018). The development of cost-effective, non-invasive biomarkers in alternative body fluids or tissues for the detection of early cognitive dysfunction remains an active research focus. Recent studies have identified blood-based biomarkers, including tau protein, neurofilament light chain (NFL), and Aβ, as potential indicators of cognitive decline in AD, although research regarding their effectiveness remains preliminary, warranting further investigation (Baiardi et al., 2022).

A comprehensive retrospective cohort study from the Taiwan National Health Insurance Program established that elderly individuals with cataracts exhibit a 1.43-fold increased risk of developing AD (Lai et al., 2014). Light deficiency caused by cataracts may influence this process by disrupting the suprachiasmatic nucleus (SCN) regulation of circadian rhythms, subsequently exacerbating age-related conditions such as depression, insomnia, and cognitive impairment (Moncaster et al., 2022). This suggests that age-related cataract (ARC), being an age-associated condition, may predispose individuals to AD. Recent research has increasingly focused on the eye, specifically the retina, as an accessible window into brain function. Anatomically and developmentally, the retina originates from pluripotent ectodermal cells of the developing diencephalon neuroectoderm and exhibits numerous structural and functional similarities with brain tissue (Lamb et al., 2007). Initial observations of ocular symptoms in patients with AD were reported by Schlotterer et al. in 1984 (Schlotterer et al., 1984). Subsequently, in 1986, Hinton et al. first documented histological evidence of retinal abnormalities in AD, including substantial loss of retinal ganglion cell neurons, decreased thickness of the retinal nerve fiber layer (RNFL), and optic nerve degeneration (Hinton et al., 1986). Given these findings, ocular biomarkers present promising opportunities for early identification of AD through non-invasive and multi-modal approaches. Therefore, this study aimed to develop and validate a prediction model to facilitate early identification of MCI in patients with ARC who demonstrate risk factors for the progression of AD.

With demographic shifts, an increasing number of individuals with low vision or blindness are experiencing cataracts (Flaxman et al., 2017). Currently, cataract surgery remains the only effective treatment, characterized by brief operative time and rapid postoperative recovery, enabling the development of ambulatory day surgery (Lawrence et al., 2015). Consequently, cataract surgery is increasingly performed in outpatient clinics rather than inpatient departments, owing to its efficacy, safety, and cost-effectiveness. However, patients with ARC are typically older and face a higher risk of concomitant cognitive dysfunction. The limited time available for communication during day surgery requires enhanced coordination. Effective management of these patients, particularly through early identification of cognitive impairment, is crucial to predict their coordination levels during the operation.

This study introduces a novel visual and user-friendly nomogram prediction model utilizing non-invasive and readily obtainable ocular biomarkers designed for the early identification of cognitive decline risk among patients with ARC. The nomogram incorporates six independent prognostic factors: age, years of education, DBP, SPCA flow rate, pRNFL (I), and vertical C/D, demonstrating significant discriminative ability in the training dataset (AUC: 0.791, 95%CI: 0.722–0.864) and the test dataset (AUC: 0.750, 95%CI: 0.627–0.85) to differentiate between individuals with MCI and those with normal cognitive function among patients with ARC, as assessed using the bootstrap method. Further validation through DCA confirmed the nomogram’s strong consistency and clinical utility.

In the training dataset, the inclusion of ocular predictors enhanced overall prediction accuracy, proving more effective than the baseline model. In the test dataset, only the IDI demonstrated a statistically significant difference, while the area under the ROC curve (AUC) and NRI showed no significant differences compared to baseline model. On the one hand, this may be related to the insufficient sample size of the test set. On the other hand, this discrepancy may be attributed to the distinct aspects emphasized by different metrics. IDI can sensitively capture continuous improvements in individual risk prediction without relying on classification thresholds, whereas the NRI measures the reclassification ability of a model under predefined risk thresholds and depends on the setting of these thresholds. Additionally, if the number of correctly upgraded cases in the event group and incorrectly upgraded cases in the non-event group offset each other, the difference in NRI might also fail to reach significance. AUC reflects overall discriminative ability but may be insensitive to localized improvements in predicted probabilities (e.g., within intermediate risk ranges). The statistical significance of IDI indicates that the new model achieved substantive improvements in the absolute accuracy of predicted probabilities or risk gradients. However, these improvements were not translated into enhanced classification capability (as reflected by NRI) or overall discriminative performance (as measured by AUC). In clinical practice, the model incorporating ocular indicators may still hold practical value for applications such as personalized risk assessment. We have reviewed pertinent literature and found that there is currently no existing model utilizing ophthalmology-related indicators to predict the risk of MCI, which highlights the innovative nature of our work. Similar type of work includes a cross-sectional study based on machine learning algorithms to predict MCI in older adults dominated by Song et al. (2025). In their study, 6,434 older adults were enrolled based on the data of the China Health and Elderly Care Longitudinal Survey (CHARLS) in 2020. Six machine learning (ML) algorithms were employed in this study: logistic regression, K-nearest neighbors (KNN), support vector machine (SVM), decision tree (DT), LightGBM, and random forest (RF). These algorithms identified five key characteristics for predicting MCI: educational level, social engagement, gender, relationship with children, and age. Some of these findings align with those of our own study. Ultimately, the area under the ROC curve for each model ranged from 0.71 to 0.77. In a study conducted by Yuan et al. focused on the early identification and support of individuals at risk of developing cognitive impairment following traumatic brain injury, several significant independent predictive factors were identified, including age, years of education, pulmonary infection status, epilepsy status, cerebrospinal fluid leakage status, and the Helsinki score (Yuan et al., 2025). Additionally, a nomogram was developed and translated into an online risk calculator, akin to the approach taken in our study.

Aging is a complex and progressive process characterized by systematic changes occurring over decades (Yin et al., 2015). Research has established that aging leads to metabolic dysregulation, insomnia, depression, and cognitive decline (Rettberg et al., 2016). Age serves as a fundamental catalyst in the development of AD (National Institute on Aging, 2015). This condition is linked to glucose hypometabolism, disrupted cholesterol homeostasis, mitochondrial dysfunction, altered immune and inflammatory responses, Aβ processing, white matter deterioration, and reduced regenerative capacity (Cai and Jeong, 2020; Klosinski et al., 2015; Masters et al., 2015; Shaw et al., 2013). This study revealed a significant positive correlation between advanced age and increased cognitive impairment risk (OR: 1.097; 95%CI: 1.041–1.161; p < 0.001), which aligned with previous research findings. Whitley et al. (2016) examined five cognitive measures in a large, representative UK population sample of over 40,000 individuals aged 16-100 years. The results demonstrated that all measured cognitive functions showed an earlier decline beginning around age 60.

Research indicates that educational attainment may serve as a protective factor against cognitive decline, commonly termed cognitive reserve. Education can increase regional cortical thickness in healthy individuals, contributing to increased brain reserve, while also enabling patients with AD to better manage brain atrophy effects through enhanced cognitive reserve (Liu et al., 2012). However, the cognitive benefits of education may differ across demographic characteristics such as gender and ethnicity (Sortsø et al., 2017). Additionally, education’s influence on cognitive function is closely linked to vascular pathology and appears most significant during early disease phases, highlighting the intricate relationship between education, brain health, and cognitive outcomes (Zieren et al., 2013). This study demonstrated the protective role of higher education in cognitive function among patients with ARC (OR: 0.333; 95%CI: 0.140–0.749; p = 0.010).

The risk of MCI showed a negative correlation with DBP (OR: 0.949; 95%CI: 0.907–0.990; p = 0.019). Longitudinal studies have supported these findings, suggesting that a mild to moderate increase in DBP could reduce the risk of developing AD (Verghese et al., 2003; Yang et al., 2011). A potential explanation for this relationship is that lower DBP may be insufficient for maintaining adequate cerebral perfusion, potentially contributing to cerebral Aβ accumulation (Li et al., 2022). During later life stages, increased arterial stiffness manifests through decreased DBP and elevated SBP. Prolonged exposure to elevated pulse pressure may lead to cerebral white matter damage, brain atrophy, and deterioration of cortical connections, subsequently affecting cognitive function (Del Pinto et al., 2021). Alternatively, this phenomenon might suggest that dementia onset could influence the central regulation of blood pressure, resulting in lower DBP (Masoli and Delgado, 2021). Further research has emphasized the connection between DBP and hippocampal volume, indicating that elevated DBP significantly correlates with increased hippocampal volume, potentially influencing cognitive health (Ngwa et al., 2018). Additionally, blood pressure variability, including DBP fluctuations, has been linked to cognitive outcomes. Studies on visit-to-visit blood pressure variability have shown that increased DBP variability correlates with higher risks of MCI and probable dementia, emphasizing the importance of BPV monitoring in clinical practice (Guo et al., 2023).

Since the initial documentation of AD pathology in the retina of patients with AD in the 1980s (Hinton et al., 1986), substantial evidence has emerged clarifying the relationship between retinal changes and AD. Advancements in ophthalmic technologies, particularly SD-OCT, have significantly improved resolution compared to time domain OCT, enabling detailed examination of all retinal layers in patients with AD. Research has identified a gradient of retinal thickness reduction, with more pronounced thinning in the inner retinal layers compared to the outer layers (Asanad et al., 2019). Studies across diverse populations have indicated a potential correlation between inner retinal thickness and cognitive function. MRI studies in non-demented individuals have revealed a possible connection between GC-IPL thickness and temporal and occipital lobe atrophy (Ong et al., 2015). Additionally, RNFL thinning has been associated with brain alterations in visual and limbic networks (Ong et al., 2015; Méndez-Gómez et al., 2018). While most studies indicate that retinal layers change progressively with AD progression, some research reports show no statistically significant differences (Asanad et al., 2019; Chan et al., 2019; Kesler et al., 2011; Knoll et al., 2016; Tao et al., 2019). These discrepancies may result from variations in exclusion criteria, cognitive assessment methodologies, and handling of confounding factors. In our univariate analysis, neither ILM-RPE thickness nor GC-IPL thickness showed significant differences between the MCI and normal groups (p > 0.05). However, pRNFL(I) thinning emerged as a risk factor for MCI (OR: 0.979; 95%CI: 0.964–0.993; p = 0.005). The findings suggest that pRNFL thinning precedes changes in other retinal layers, potentially indicating neurodegeneration of the central nervous system. Postmortem analyses of AD retinas have demonstrated cellular shrinkage, swelling, and vacuolization (Cheung et al., 2017; Dehghani et al., 2018). Optic disk pallor was observed even in early AD stages, attributed to axonal loss and perfusion alterations (Bambo et al., 2015). A meta-analysis indicated that a higher cup-to-disc ratio, lower height variation contour, lower rim area, and lower rim volume measured by scanning laser ophthalmoscopy may facilitate the diagnosis of AD (Ge et al., 2021). Our OCT study findings confirmed an enlarged cup-to-disc ratio as a significant risk factor for MCI (OR: 11.927; 95%CI: 1.059–155.308; p = 0.049).

The retinal vasculature shares structural and functional similarities with the cerebral vasculature. Research indicates that alterations in blood flow parameters precede neuronal loss (Javaid et al., 2016). This finding suggests that changes in retinal blood vessels could reflect underlying cerebrovascular pathology. The ophthalmic artery, originating from the internal carotid artery, supplies blood to the ocular region and gives rise to the central retinal artery and posterior ciliary arteries. The short posterior ciliary artery branches within the choroid, forming a choroidal vascular network that supplies the choroid, macula, and the outer retinal layers, while the central retinal artery delivers blood to the inner retina (Baldoncini et al., 2019). In this study, patients with ARC demonstrated an increased SPCA flow rate, which correlated with a higher likelihood of developing cognitive impairment (OR: 1.063; 95%CI: 1.008–1.132; p = 0.038). Furthermore, OCTA analysis of the superficial capillary plexus revealed no significant differences in SCP-VP and SCP-DP of the macular and optic disc regions between the MCI group and the normal group (p > 0.05). A study by Yoon et al. comparing patients with MCI to those with normal cognitive function found no significant differences in VD, PD, vein diameters, and the area of the foveal avascular zone (FAZ) (Yoon et al., 2019), aligning with our findings. Diverse perspectives exist on this matter (Chua et al., 2020). Based on our findings, we hypothesize that during the progression from cognitive normality to early cognitive impairments, changes in blood flow parameters of ocular supply vessels precede alterations in the retinal capillary network and retinal structure. These modifications in ocular blood flow initially arise from feedback mechanisms, such as increased arterial flow velocity. Current research exploring the relationship between ocular arteries and cognitive function remains limited. However, as components of the systemic vasculature, ocular arteries have been implicated in optic nerve damage in glaucoma, associated with changes in ocular artery blood flow and hypertension (Waliszek-Iwanicka et al., 2010). This suggests that alterations in ocular blood flow may impair visual function through effects on the optic nerve (e.g., enlarged cup-to-disc ratio), potentially contributing to cognitive-related issues.

Although recent research strongly suggests a correlation between ocular biomarkers and cognitive impairment, no definitive diagnostic model has been established. This study introduces a new nomogram prediction model based on six indicators obtained through non-invasive, accessible, safe, and cost-effective methods. The nomogram demonstrates robust performance in internal and external verification, making it applicable to diverse patient populations. Additionally, our study provides a visual representation beneficial for clinicians, particularly in basic hospitals with limited resources. This study has several limitations that warrant consideration. First, owing to the heterogeneity of OCT/OCTA devices in clinical practice, variations in image analysis software may affect readouts. Hence, this calculator applies only in settings using the OptoVue device. Second, certain potential confounding factors (such as ApoE genotype) were not fully incorporated, potentially affecting the final results. Third, the limited sample size may affect the generalizability of the results. Furthermore, the cross-sectional study design precludes causal inferences. Future large-scale cohort studies are necessary to validate these findings and conclusions.

This research culminates in the development of an online calculator that enables the identification of early cognitive dysfunction in patients with ARC through ocular biomarkers. The validation results confirm the model’s robust discrimination ability. The identification of novel ocular biomarkers represents a significant advancement in the assessment and management of cognitive impairment for both patients and healthcare providers.