Patients with LLD were recruited from the Department of Beijing Huilongguan Clinical Medical School, Peking University, PR, China. All patients met the following inclusion criteria: (1) 60–85 years old; (2) newly admitted, met the criteria of the Diagnostic and Statistical Manual of Mental Disorders, V Edition (DSM-5) for unipolar MDD, had a score >17 in the 17-item Hamilton Depression Rating Scale (HAMD-17), and Clinical Dementia Rating (CDR) <0.5; (3) received no antidepressant therapy for at least 2 weeks; (4) had at least 5 years of formal education; and (5) Han ethnicity and right-handedness. The exclusion criteria were as follows: (1) history of other psychiatric disorders before depression onset; (2) history of severe medical illness, severe brain injury, epilepsy, cerebrovascular disease, or long-term neuropsychiatric complications such as tumors; (3) history of head trauma, Parkinson’s disease, or multiple sclerosis; and (4) alcohol or drug abuse. Healthy volunteers were recruited from local communities through advertisements. The inclusion criteria for all healthy controls (HCs) were as follows: (1) No serious mental illness according to the DSM-5 criteria; (2) no severe medical illness; (3) no history of neurological disorders or MRI evidence of structural brain abnormalities; and (4) no family history of psychiatric disorders.

The Ethics Committee of Beijing Hui Long Guan Hospital approved this study. Informed consent was obtained from all patients.

Depressive symptoms were measured using the HAMD-17 score, and the Hamilton Anxiety Scale (HAMA) was used to assess anxiety and depression symptoms.

The MMSE, a widely used screening instrument administered by trained researchers, provides a rapid initial assessment of global cognitive function. It evaluates cognitive domains through 19 items, with a total score of 30 points. A score of ≥26 indicates normal cognition, while that of <26 suggests cognitive impairment. The Repeatable Battery for the Assessment of Neuropsychological Status (RBANS), a standardized measurement tool that includes five subscales (immediate memory, delayed memory, attention, language, and visuospatial function), has good internal consistency and acceptable reliability in Chinese patients with MDD (23). Therefore, the RBANS was used to identify and characterize cognitive decline in patients with depressive disorders.

Venous blood (4 mL) was collected from each patient between 06:00 AM and 07:00 AM after 12 h of fasting. Hcy concentrations were quantified enzymatically on a Beckman Coulter AU5800 analyzer. Reagent blank absorbance exceeded 0.80; intra-assay CV (10 replicates) was ≤5%; analytical sensitivity (absolute ΔA at 10 μmol/L) was >0.2.

SPSS 22.0 software (SPSS Inc., Chicago, IL, USA) was used to analyze demographic variables, including age, body mass index, and sex. Qualitative variables are presented as frequencies (counts) and percentages, while quantitative variables are reported as means ± standard deviations. Chi-squared tests compared dichotomous variables between groups, and independent sample t-tests or one-way ANCOVAs analyzed continuous variables. Bonferroni correction adjusted for multiple comparisons across RBANS subscales.

Pearson’s correlation analysis examined the relationship between Hcy levels and clinical parameters. A multiple linear regression model with interaction terms assessed whether HAMD modified the effect of Hcy on cognitive function (total RBANS score). The model included main effects of Hcy, HAMD, and their interaction (Hcy × HAMD), with age, education, and gender adjusted as covariates. Collinearity was evaluated via variance inflation factors, and interaction term significance was tested using t-statistics. Mediation analysis via Bootstrap resampling (5,000 iterations) explored the mechanism between Hcy and MMSE using “Immediate Memory” and “Delayed Memory” score grouping as the mediator. Statistical significance was defined as p < 0.05.

Sixty patients with LLD and 46 HCs were eligible for inclusion. The demographic and clinical characteristics of the participants are presented in Table 1 . No significant differences were observed between the LLD and HC groups regarding age, sex and other demographic characteristics (all p > 0.05) ( Table 1 ).

The mean MMSE total score in patients with depression was lower than that in the HC group (p < 0.001). Overall, 60% of the patients with depression had cognitive impairment, defined as an MMSE score <26. Compared with the HCs, all patients showed some cognitive deficits in the overall RBANS subdomains, except for the language score (all p < 0.05) ( Table 2 ).

The patients with LLD were further divided into two groups according to their MMSE scores. The first group included patients with a low MMSE score (<26; L-MMSE) and second group included those with a normal MMSE score (≥26; N-MMSE). No significant differences in age, sex, height, weight, or HAMD-17 or HAMA scores were observed between the two groups (p > 0.05) ( Table 1 ).

Compared to the HCs, significantly worse performance was observed in terms of immediate memory, attention, delayed memory, and the total scale score in RBANS in the N-MMSE group (p < 0.05). In addition, significantly greater cognitive impairment was observed in the L-MMSE group compared to both HC and N-MMSE groups across all domains (all p < 0.05; subscale-specific differences detailed in Table 2 ).

Elevated Hcy levels were observed in patients with LLD (F = 8.560, p < 0.001) ( Table 1 ), which were dependent on group allocation per MMSE score. In particular, Hcy levels in patients in the L-MMSE group were higher than those in N-MMSE group and HCs (p < 0.05). No significant differences in Hcy levels were observed between patients in the N-MMSE group and HCs (p > 0.05) ( Table 1 ; Figure 1 ).

In the patients with LLD, Hcy levels were significantly negatively associated with MMSE scores(r = -0.286, confidence interval [CI]: -0.496–0.073, p = 0.029), immediate memory (r = –0.283, 95%CI: -0.497–0.106, p = 0.031), delayed memory (r = -0.279, 95%CI: -0.480–0.059, p = 0.003), and the total RBANS scale score (r = −0.270, 95%CI: -0.470–0.022, p = 0.034). No significant correlation was observed between Hcy levels and cognitive function in the HC group (all p > 0.05) ( Figure 2 ). We further performed multiple linear regression analysis to adjust for confounding variables such as age, sex, and education. The results showed that after adjusting for confounding factors, serum Hcy levels were independently and negatively correlated with RBANS scores (β = -0.311, 95%CI: -4.310 to -0.310, p = 0.024). Further inclusion of an interaction term between Hcy and HAMD significantly improved the model’s goodness of fit (ΔR² = 0.058, p = 0.019). The interaction effect test revealed that the impact of Hcy on cognitive function was increased with the severity of depression (β = 1.385, 95%CI: 0.006–0.589, p = 0.046) ( Table 3 ).

In the subgroup analysis, Hcy levels in the N-MMSE group were negatively associated with MMSE scores (r = -0.587, p = 0.003), attention (r = -0.415, p = 0.049), delayed memory (r = -0.419, p = 0.046), and the total RBANS scale score (r = -0.499, p = 0.015). No significant associations were observed in the L-MMSE group (all p > 0.05) ( Figure 3 ). Further, we conducted mediation analysis using immediate memory and delayed memory as mediating variables, and found that the direct effect of Hcy on MMSE was nonsignificant (t = -1.221, p = 0.227), with Hcy significantly predicting immediate memory (t = -2.174, p = 0.038), and delayed memory (t = 4.593, p < 0.001) significantly predicting MMSE, and the Bootstrap-derived indirect effect of -0.192 (95%CI: -0.479, -0.052) and -0.152 (95%CI: -0.362, -0.035) excluding zero, indicating full mediation by immediate memory and delayed memory ( Figure 4 ).

In this study, the cognitive performance—including immediate memory, visuospatial function, attention, and delayed memory—in patients with LLD was lower than that in healthy older individuals. This finding confirms previous reports on the presence of mild-to-moderate cognitive dysfunction in patients with LLD (6, 8, 9, 24). For example, a previous study reported that individuals with depression exhibited significant attention impairments, with reduced ability to sustain focus and filter out distractions (4). Memory deficits, particularly regarding episodic and working memory, have also been frequently reported (24). Older adults with depression performed poorly in recalling recent events and manipulating information in short-term memory (8). Executive function, which encompasses skills such as planning, problem-solving, and cognitive flexibility, is also compromised in patients with LLD. Furthermore, older adults with depression struggle with tasks that require strategic decision-making and organization of complex information (6). Paelecke-Habermann (25) summarized attention, executive function, and memory deficits after remission, and found that reduced depressive symptoms did not necessarily lead to cognitive improvement, possibly because cognitive impairment is a trait marker rather than a state marker of depression. These impairments may be attributed to several factors. One possible mechanism underlying cognitive impairment in LLD is the dysregulation of the system of Hcy metabolism (21). Changes in Hcy levels have been implicated in depressive symptoms and cognitive dysfunction (20, 21).

In this study, Hcy levels were elevated in older patients with depression, consistent with the results of previous research. Older adults with depression have higher Hcy levels than their healthy counterparts (13). Large meta-analyses have shown increased serum and plasma Hcy levels in patients with MDD compared to healthy individuals (13). Compared with MDD, LLD is characterized by less prominent affective symptoms, obvious somatic symptoms, and more severe cognitive impairment (2, 8, 9). Whether regarded as a special subtype of MDD or an independent diagnostic entity, LLD has been considered in previous studies to have specificity in its pathogenesis and related biological manifestations (6). For example, some studies suggest that the role of Hcy in the pathogenesis of MDD may be related to metabolites of monoamine neurotransmitters, including serotonin, dopamine, and norepinephrine (26), while the association between Hcy and the pathogenesis of LLD is more influenced by organic factors such as cerebrovascular conditions. As shown in some studies, elevated Hcy levels have been associated with vascular endothelial dysfunction and increased oxidative stress, both of which contribute to neuroinflammation and neuronal damage, thereby influencing mood regulation (27, 28).

The current study specifically highlights the connection between elevated Hcy levels and compromised cognitive function in patients with LLD. Cognitive performance and Hcy levels differed in the L-MMSE and N-MMSE groups. Notably, Hcy levels were significantly increased in the L-MMSE group but correlated with cognitive impairment in the N-MMSE group.

This seemingly paradoxical phenomenon may be related to a “threshold-dependent neurotoxicity mechanism” of Hcy. The role of NMDA receptors in cognitive impairment is well established[]. Research indicates that Hcy has different effects on NMDA receptors depending on its concentration (29). Studies by Bolton et al. showed that Hcy within the physiological concentration range increases the peak current of NMDA receptors, reduces receptor desensitization, and enhances glutamatergic transmission (30). Furthermore, Hcy can act as an agonist or co-agonist at NMDA receptors, increasing the excitability of glutamatergic neurons (30). In contrast, at higher concentrations (>15 μmol/L), Hcy can rapidly amplify its own concentration through an “auto-sensitization response.” High concentrations of Hcy exert neurotoxic effects. It interacts with the glutamate/cystine transporter (also known as the Xc- transport system), leading to activation of adjacent ionotropic glutamate receptors. This causes excessive NMDA receptor activation, resulting in excitotoxicity, disruption of synaptic formation, loss of prepulse inhibition, and ultimately, neuronal death (31). At this stage, neuronal compensatory mechanisms fail. This may explain why the linear association disappears in patients with severe cognitive impairment (L-MMSE group). It also indicates that the same metabolic indicators may exert distinct effects on diseases, depending on the underlying disease state. As shown in a large-sample study: analysis of NHANES data revealed a positive correlation between HDL levels and cognitive function, yet this protective effect of HDL on cognition was absent in individuals with depressive symptoms (32). This phenomenon further underscores the complex interplay between physiological indicators and disease processes.