As of December 31, 2024, we systematically searched five databases (PubMed, Embase, Web of Science, Google Scholar, and Elsevier) for articles published from January 1, 2014, to December 31, 2024. The search strategy combined the following terms using Boolean operators: (Diabetes Mellitus, Type 2 OR T2DM OR type 2 diabetes) AND (Cognitive Dysfunction OR Mild Cognitive Impairment OR MCI) AND (risk factors OR predictors OR determinants). In addition, manual searches of reference lists from identified articles and relevant reviews were performed to supplement the electronic search.

One researcher (H.L.W.) performed the initial literature search and removed duplicates. Two researchers (Y.Z. and G.H.T.) independently screened titles and abstracts against predefined inclusion and exclusion criteria. Full texts were retrieved if either reviewer deemed an article potentially eligible. The reviewers then independently assessed the full-text articles for final inclusion. Discrepancies were resolved through consultation with the corresponding author (R.L.).

The inclusion criteria were as follows (1): study involving patients diagnosed with T2DM and MCI (2); case-control, cohort, or cross-sectional designs (3); data convertible to odds ratio (OR) with a 95% confidence interval (CI) (4); reporting at least one risk factor (5); use of multivariable logistic regression to identify T2DM-MCI determinants; and (6) clear diagnostic criteria for MCI. The exclusion criteria were (1): duplicate publications (2); reviews, letters, or non-research articles; and (3) non-English publications.

Data extraction was independently performed by two investigators (Y.Z. and L.L.W.) using standardized forms. The following parameters were recorded: first author’s name, mean age, sex distribution, publication year, study location, sample size, prevalence of T2DM-MCI comorbidity, and reported risk factors. Quantitative measures, including OR with corresponding 95%CI, were extracted for each determinant. The extracted variables were stratified into four etiological domains: 1) Demographic and Clinical Characteristics: age, sex, diabetes duration, body mass index (BMI), and educational attainment. 2) Lifestyle Factors: alcohol consumption, smoking status. 3) Comorbidity Management: depression, hypertension, cardiovascular disease (CVD), and diabetic retinopathy (DR). 4) Biochemical Indicators: HbA1c, homeostasis model assessment of insulin resistance (HOMA-IR), low-density lipoprotein cholesterol (LDL-C), high-sensitivity C-reactive protein (HS-CRP), FPG, and high-density lipoprotein (HDL).

Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for cohort and case-control studies (33) and the Joanna Briggs Institute (JBI) Critical Appraisal Tool for cross-sectional studies (34). The NOS evaluates three domains: selection, comparability, and exposure/outcome ascertainment, with a maximum score of 9 points. Studies scoring ≤ 5 points were classified as low quality, scoring between 5–7 points were classified as medium quality, and scoring > 8 were classified as high quality. The JBI tool employs a percentage-based scoring system, with a maximum score of 8 points, categorizing studies as high quality (≥ 7), moderate quality (5, 6), or low quality (≤4). Only studies meeting quality thresholds (NOS ≥ 5 or JBI ≥5) were retained. Two investigators (G.H.T. and H.L.W.) independently conducted quality assessments. Discrepancies in scoring were resolved through consultation with the corresponding author (R.L.).

All statistical analyses were performed using Review Manager (RevMan) software, version 5.4. OR with corresponding 95% CI served as effect size measures. Heterogeneity was quantified using I² statistics and P-values, with thresholds set at P < 0.1 or I² > 50% indicating substantial heterogeneity. The fixed-effects model assumes consistent effect sizes across studies, suitable for low heterogeneity (I² ≤50%) and calculates pooled effect size through weighted averages. The random-effects model assumes variability in effect sizes and is used when significant heterogeneity exists (I² >50%), incorporating study differences through weighted averages. Therefore, a random-effects model is chosen when I² >50%, and a fixed-effects model is used otherwise. Subgroup analyses were conducted for factors with ≥ 10 studies. Sensitivity analyses were performed by switching between fixed-effects and random-effects models for outcomes demonstrating I² > 50%. Publication bias was assessed through funnel plot symmetry evaluation and Egger’s linear regression test, which was restricted to factors with ≥10 studies. A significance level of P < 0.05 was defined for all inferential analyses.

The systematic search initially identified 4,689 citations. Following duplicate removal, 3,326 records were subjected to preliminary screening. Title/abstract screening excluded 2,956 non-eligible studies, leaving 212 articles for full-text assessment. Ultimately, 30 studies met the inclusion criteria and were included in the meta-analysis of T2DM-MCI determinants. The complete screening protocol is presented in the PRISMA flowchart ( Figure 1 ).

Table 1 summarizes the characteristics of the included studies. The pooled analysis comprised 30 studies conducted across five countries, involving 10,469 participants with T2DM, including 4,516 cases with comorbid MCI (prevalence rate = 44.1%). These investigations, published between 2014 and 2024, had sample sizes ranging from 103 to 1,278 participants. The mean age of the study populations ranged from 50 to 84 years, with female participants representing 41.8% of the total cohort.

The included studies were of moderate to high quality. Overall quality assessment showed that 8 articles (35–42) were classified as high quality, while 22 articles (6, 26, 27, 43–61) were categorized as medium quality, For specific scores, see Supplementary Table S1 and Supplementary Table S2 . The included studies demonstrated stability in sensitivity analyses. Funnel plots indicated no significant publication bias ( Figure 2 ). Table 2 presents the results of Egger’s test, confirming no substantial publication bias. For the sensitivity analysis, see Supplementary Table S3 .

Subgroup analysis in this study demonstrated that T2DM patients aged ≥ 70 years face an elevated MCI risk (OR = 1.06, 95% CI: 1.04-1.08, P < 0.001), while those with diabetes aged 60–69 years showed a higher risk (OR = 1.07, 95%CI:1.01-1.14, P = 0.005). These findings align with Sun et al. (63), confirming age as a non-modifiable risk factor for MCI. Potential mechanisms include age-related neurodegeneration (e.g., neuronal loss, reduced synaptic density, and impaired cerebral energy metabolism) (64), compounded by elevated advanced glycation end products (AGEs) and oxidative stress in elderly T2DM patients (65), which synergistically accelerate vascular dysfunction and cognitive deterioration (66).

Female sex was identified as a significant risk factor for MCI in T2DM patients (OR = 1.48 vs. 0.91 in men), consistent with You et al. (67). While gender differences remain debated (68, 69), emerging evidence suggests that the postmenopausal decline in estrogen may attenuate neuroprotective pathways (68, 70). Additionally, sex-specific disparities in cardiovascular risks profiles (71), adiposity distribution patterns (72), and chronic inflammatory states (73) likely contribute to this association. Although these mechanisms are not fully understood, existing studies support the significant role of sex in diabetes-related cognitive impairment.

The present study highlights a strong relationship between diabetes duration and MCI risk. Subgroup analyses revealed significantly elevated risks in patients with longer disease duration: those with ≥ 10 years of diabetes exhibited an adjusted OR of 1.04 (95%CI:1.01-1.07, P = 0.02), while those with 8–9 years diabetes showed a higher risk (OR = 2.56, 95%CI:2.13-3.08, P < 0.001). These findings suggest that prolonged hyperglycemia accelerates cognitive dysfunction through cumulative metabolic insults. Longitudinal studies indicate that patients with ≥ 20 years of diabetes have a 3.32-fold increased risk of information processing deficits, a 1.72-fold risk of immediate recall impairment, and a 1.76-fold risk of executive dysfunction compared to those with shorter disease duration (74). Mechanistically, chronic hyperglycemia drives insulin resistance, intermittent hypoglycemia, and microvascular complications (75). Furthermore, extended disease duration may induce structural and functional brain changes (e.g. accelerating cerebral atrophy and reduced synaptic density) and functional neurodegeneration (76), which collectively contribute to cognitive decline.

Low educational attainment (≤ 6 years) was associated with an increased risk of MCI (OR = 1.29, 95% CI: 0.80–0.97, P = 0.007), supporting the cognitive reserve hypothesis (77). Higher educational attainment may enhance neural plasticity and compensatory mechanisms, potentially delaying cognitive decline. Additionally, greater education may optimize neural network efficiency, helping maintain cognitive resilience despite chronic metabolic conditions such as diabetes (36).

No significant association was found between BMI and MCI risk in patients with T2DM (OR = 1.18, 95% CI: 0.94–1.49, P = 0.15). Although elevated BMI is associated with insulin resistance (78), adiposity-related inflammation (79), and cardiovascular risks (80)—factors that may indirectly impair cerebral metabolism and cognition— BMI was not independently associated with MCI in this cohort. This finding may reflect interactions between BMI and confounding variables (e.g., age, glycemic control, sex). Future research should clarify BMI’s role through stratified analyses and longitudinal studies.

It was demonstrated that smoking increases the MCI risk in patients with T2DM (OR = 1.44, 95% CI: 1.18–1.75, P = 0.0003). These findings align with previous studies (81, 82), confirming smoking as a critical risk factor for T2DM-MCI comorbidity. For instance, Hagger-Johnson et al. (83) reported accelerated cognitive decline in middle-aged and elderly smokers, while Xia et al. (84) identified an inverse correlation between smoking intensity and serum brain-derived neurotrophic factor levels—a key mediator of neurogenesis and synaptic plasticity. Mechanistically, nicotine may impair cognition through interactions with nicotinic acetylcholine receptor subunits (e.g., α4, β2, and α7) [ (85). Although precise pathways require further elucidation, the robust association between smoking and cognitive deterioration is well-established.

In contrast, alcohol consumption showed no significant association with T2DM-MCI comorbidity in this study (OR = 0.81, 95% CI: 0.36–1.80, P = 0.61). However, existing evidence (86) suggests a complex dose-response dynamic, including U-shaped or J-shaped relationships. A Finnish cohort study of 1,464 adults aged 65–79 years found that midlife heavy drinking patterns significantly increased MCI risk (OR = 5.08, P = 0.020) (87), while another study reported elevated cognitive decline risks in both heavy drinkers (OR = 1.44, 95% CI: 1.02–2.10) and abstainers (OR = 1.94, 95% CI: 1.10–3.44) compared to moderate drinkers (88). The non-significant association observed here may reflect limitations of the cross-sectional design or population heterogeneity, necessitating longitudinal studies to clarify the role of alcohol in T2DM-related cognitive dysfunction.

Hypertension was identified as a significant risk factor for MCI in patients with T2DM (OR = 2.25, 95% CI: 1.49–3.40, P < 0.001), consistent with prior research (89, 90). Chronic hypertension induces structural and functional cerebrovascular damage through ischemic white matter injury and microvascular pathology, reducing cerebral blood flow and accelerating cognitive decline (91). Similarly, CVD significantly elevates MCI risk (OR = 2.61, 95% CI: 1.99–3.43, P < 0.001), corroborating Xie et al. (92). Post-stroke cerebrovascular injuries—particularly those involving extracranial carotid or intracranial vascular lesions (93)—are strongly associated with cognitive impairment in diabetic populations. Notably, diabetic stroke survivors with larger infarct volumes exhibit pronounced cognitive deficits, substantially increasing post-stroke cognitive impairment (PSCI) risk (94). These findings underscore the need for intensified CVD management in T2DM patients to mitigate cognitive deterioration.

Interestingly, depression showed no significant association with MCI risk (OR = 2.04, 95% CI: 0.42–9.79, P = 0.38), contrasting with Carr et al. (95) and Chow et al. (96). While depression-related neurodegeneration in brain regions such as the hippocampus and prefrontal cortex may drive cognitive dysfunction (97), confounding factors [e.g. glycemic control (98), systemic inflammation (99)] and regional population differences likely explain this discrepancy.

DR emerged as a significant MCI predictor (OR = 1.50, 95% CI: 1.12–2.01, P = 0.006), corroborating Gorska-Ciebiada et al. (45). As a microvascular complication, DR shares pathophysiological mechanisms with cerebral microangiopathy [e.g. chronic hyperglycemia-induced endothelial dysfunction (100)], suggesting its potential role as a biomarker for concurrent brain microvascular damage (101). Proactive DR screening and management may thus help reduce MCI risk in T2DM patients.

It was found that HbA1c >9% significantly elevates MCI risk in patients with T2DM (OR = 1.33, P = 0.001), with residual risks persisting even within the conventional glycemic target range (HbA1c <7%: OR = 1.38, P = 0.002). Chronic hyperglycemia impairs glial cell function, inducing cerebrovascular pathology and neuronal damage that ultimately compromise cognition (102). Zheng et al. (103) reported that each 1 mmol/mol increase in HbA1c exacerbates declines in cognitive, memory, and executive functions. Hyperglycemia-driven cognitive deterioration involves multiple mechanisms, including disrupted neurotransmitter metabolism (104), aggravated oxidative stress (105), and altered neuronal energy homeostasis (106), solidifying HbA1c’s role as a critical biomarker for T2DM-MCI comorbidity.

Elevated HOMA-IR was independently associated with MCI risk (OR = 1.95, 95% CI: 1.14–3.35, P = 0.02). As a surrogate marker of insulin resistance, HOMA-IR reflects compensatory hyperinsulinemia, which is closely linked to cognitive decline. Kim et al. (107) identified strong associations between hyperinsulinemia and impairments in memory and executive function. Pharmacological interventions improving insulin sensitivity have shown potential to enhance memory performance (108), suggesting therapeutic relevance. Although insulin resistance-MCI relationships are empirically supported (108, 109), underlying mechanisms remain incompletely elucidated, necessitating further investigation of HOMA-IR’s predictive utility.

A potential association between elevated FPG and MCI risk in T2DM patients was also identified (OR = 1.15, 95% CI: 1.01–1.32, P = 0.04). Comparative analysis revealed lower FPG levels in non-MCI groups, suggesting that sustained hyperglycemia may contribute to cognitive dysfunction. Chronic hyperglycemia not only exacerbates microvascular complications but also induces long-term neurological detriment, oxidative stress, and blood-brain barrier disruption (110, 111). Thus, FPG serves dual roles as a glycemic control biomarker and a predictor of cognitive decline in T2DM.

HDL levels showed no significant association with MCI risk in T2DM patients (OR = 1.07, 95% CI: 0.79–1.43, P = 0.68), diverging from studies suggesting HDL’s neuroprotective effects via anti-inflammatory and antioxidant pathways (112, 113). Moderate heterogeneity across studies (I² = 65%, P = 0.06) may stem from methodological variations in HDL measurement or population characteristics.

No significant association was found between LDL-C levels and MCI risk in the current study (OR = 0.99, 95% CI: 0.52–1.89, P = 0.99). Despite proposed U-shaped relationships between LDL-C and cognitive function (114), significant heterogeneity (I² = 80%, P = 0.006) suggests confounding effects from diabetes control status or inflammatory mediators, necessitating further investigation into these complex interactions.

Finally, elevated HS-CRP was strongly associated with MCI risk (OR = 2.85, 95% CI: 2.09–3.89, P <0.001). As a sensitive inflammatory biomarker (115), elevated HS-CRP reflects systemic inflammation that synergizes with oxidative stress to amplify free radical generation, damaging cellular membranes, DNA, and neuronal function (116, 117). These findings highlight HS-CRP’s dual roles as an inflammatory marker and a predictor of early cognitive decline in T2DM.