An integrated CSF-serum biomarker model for predicting clinical progression in Alzheimer’s disease

Background: The early and accurate identification of Alzheimer’s disease (AD) remains a significant clinical challenge. Integrating novel peripheral blood-based biomarkers with established cerebrospinal fluid (CSF) measures may offer a promising strategy to enhance diagnostic accuracy and risk stratification.

Methods: This study enrolled 91 participants who underwent CSF and serum testing. The cohort was randomly divided into a training set (n = 63) and an internal testing set (n = 28). External validation was performed using matched data (n = 30) from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (total n = 639). Data collected included demographics, Mini-Mental State Examination (MMSE) total scores, the Functional Activities Questionnaire (FAQ) total scores, CSF phosphorylated tau (pTau181) and amyloid-β (Aβ42) levels, and serum indices such as the albumin-to-globulin (A/G) ratio and platelet-to-lymphocyte ratio (PLR). Predictor selection was performed via univariate and multivariate logistic regression, and a nomogram was developed from the final model. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves with mean absolute error (MAE), and decision curve analysis (DCA).

Results: The final predictive model incorporated CSF pTau181, A/G ratio, and PLR (using a cut-off ≥113.22). It demonstrated robust discrimination, achieving an AUC of 0.92 in the training set, 0.86 in the testing set, and 0.83 upon external validation. Calibration was excellent (MAE = 0.039). In the testing set, sensitivity was 0.83 and specificity was 0.86. A higher A/G ratio was associated with a reduced risk of AD progression, whereas a higher PLR was associated with an increased risk.

Conclusion: The combined CSF-peripheral blood biomarker model demonstrates robust discrimination and calibration for predicting AD progression. By linking central tau pathology with peripheral nutritional and inflammatory status, it may aid clinical risk stratification and guide management strategies focused on nutrition and inflammation. Further large-scale, prospective validation is warranted.

1 Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by a gradual decline in cognitive function, frequently accompanied by neuropsychiatric symptoms in its later stages (McDade, 2022). Although the precise pathogenesis of AD remains incompletely understood (Tatulian, 2022), its biological definition has become increasingly well-established. Hallmark neuropathological changes can now be detected in vivo using cerebrospinal fluid (CSF), blood, or neuroimaging biomarkers, enabling a biological diagnosis of AD based on a positive biomarker profile, irrespective of the presence or severity of clinical symptoms (Jack et al., 2018; Jack et al., 2024).

In 2024, the National Institute on Aging-Alzheimer’s Association (NIA-AA) revised its diagnostic and staging criteria, emphasizing the biological foundation of AD diagnosis and introducing a six-stage numeric clinical staging system to more precisely characterize the severity of cognitive and functional impairments along the AD continuum. This updated framework redefines previously used concepts, such as a “prodrome of AD” or “at risk for AD, “into more clinically applicable categories: Clinical Stage 2, denoting mild detectable change with minimal functional impact, and Stage 3, representing mild cognitive impairment (MCI) with early functional limitations (Jack et al., 2024).

Accumulating evidence suggests that AD pathophysiology may commence decades before overt clinical symptoms appear (Guzman-Martinez et al., 2021). Therefore, early detection of individuals in Clinical Stages 1–2 (asymptomatic or minimally symptomatic) and Stage 3 (MCI) is critical for timely intervention (Petersen et al., 2014). With the recent approval and widespread implementation of disease-modifying therapies (DMTs), such as anti-Aβ monoclonal antibodies like lecanemab, accurately detecting these early-stage individuals and initiating DMTs promptly have become essential strategies for slowing disease progression and preserving cognitive function (Tahami Monfared et al., 2022; Cohen et al., 2023; Tahami Monfared et al., 2023). Recent studies further indicate that biomarker-based predictive models can aid in the early identification of high-risk individuals, supporting interventions targeting modifiable risk factors, and offering new prospects for primary prevention and early intervention in AD (Graff-Radford et al., 2021; van Oostveen and de Lange, 2021).

CSF biomarkers, including amyloid-beta (Aβ42), total tau (t-Tau), and phosphorylated tau (pTau181), are central to the biological diagnosis of AD (Jack et al., 2018; Jack et al., 2024). In particular, CSF and plasma pTau181 have emerged as highly specific indicators of AD-related tau pathology. Critically, elevated levels of pTau181, irrespective of the specific assay platform used, have been consistently and strongly associated with the presence and severity of AD across numerous independent cohorts and studies (Janelidze et al., 2020; Lv et al., 2023). However, the absolute value of CSF pTau181 in an individual patient does not exhibit a perfect linear relationship with current symptom severity, especially in cases with comorbidities such as vascular dementia, where clinical manifestations may be influenced by mixed pathologies. This complexity limits the utility of pTau181 alone for assessing AD severity. For these reasons, the 2024 NIA-AA criteria introduced three additional biomarker categories: inflammatory/immune mechanisms (I), vascular brain injury (V), and α-synuclein (S) (Jack et al., 2024). Among these, the “I” category includes biomarkers such as glial fibrillary acidic protein (GFAP), which reflects non-specific inflammatory processes involved in AD pathophysiology. Nevertheless, the relatively high cost of GFAP testing limits its accessibility, underscoring a pressing need to develop complementary, cost-effective, and non-invasive tools to support clinical staging and prognostic assessment (Benedet et al., 2021; Pereira et al., 2021; Jack et al., 2024).

In recent years, peripheral inflammatory markers have garnered increasing attention in AD research (Koyama et al., 2013). Accumulating evidence suggests that systemic inflammation, metabolic dysregulation, and nutritional deficiencies may influence both disease progression and clinical expression of AD and cognitive impairment (Koyama et al., 2016; Maeda et al., 2019; Zhou et al., 2021; Wang et al., 2024; Yang et al., 2024). Such readily obtainable serum biomarkers could thus serve as a practical and affordable adjunct to CSF analysis. We hypothesize that combining CSF pTau181 with serum biomarkers will yield a robust and clinically applicable model for distinguishing between early (Stages 1–2) and later (Stages 3–6) clinical stages of AD. Therefore, this study aims to develop and validate a novel combinatorial model to improve the accuracy of predicting clinical progression (Koyama et al., 2013; van der Lee et al., 2018).

Notably, nomograms have been widely used as precise instruments for predicting disease outcomes in AD and other disorders (Chen et al., 2025). These graphical tools, derived from regression models, estimate the probability of a clinical event based on multiple predictive variables and are increasingly regarded as a potential alternative or even a new standard compared to traditional staging systems (Balachandran et al., 2015; Lv et al., 2022). In this study, we aimed to develop and validate a nomogram to distinguish AD patients with objective cognitive impairment (Clinical Stages 3–6) from those without (Clinical Stage 1–2 AD).

2 Methods

2.1 Study design and participants

This study was conducted following the guidelines provided in the 2024 edition of “Developing Clinical Prediction Models: A Step-by-Step Guide” (Efthimiou et al., 2024), which outlines methodological standards for developing and validating diagnostic or prognostic prediction models. We retrospectively enrolled 103 patients who underwent lumbar puncture for cognitive assessment and CSF AD biomarker analysis at the Department of Neurology, Shenzhen People’s Hospital between 2019 and 2025. After excluding 12 patients due to incomplete data—specifically missing Mini-Mental State Examination (MMSE) scores or serum biomarker measurements—the final analytical cohort comprised 91 participants. Consistent with contemporary recommendations for feasible operationalization in clinical research (Selma-Gonzalez et al., 2025), the Mini-Mental State Examination (MMSE) was adopted as the principal criterion for cognitive staging. Functional status was concurrently assessed across all participants using the Functional Activities Questionnaire (FAQ), a useful method to discriminate and predict progression from clinical normal to MCI (Marshall et al., 2015). The resulting FAQ scores served to both characterize and provide concurrent validation that the MMSE-based stratification effectively captured the anticipated gradient in functional impairment—a dimension central to distinguishing these clinical stages.

In accordance with the 2024 NIA-AA clinical staging framework (Jack et al., 2024), participants were categorized into two distinct groups: Clinical Stages 1–2 (n = 16), defined by subjective cognitive decline alongside preserved objective cognitive function (MMSE ≥ 27) and an FAQ score ≤5. Clinical Stages 3–6 (n = 75), encompassing individuals with objectively confirmed cognitive impairment spanning mild deficits to severe dementia (MMSE ≤ 26), accompanied by an FAQ score >5. The dataset was randomly divided into a training set (n = 63) and a testing set (n = 28) at a 7:3 ratio for model development and internal validation, respectively. The overall study design is schematically presented in Figure 1.

Figure 1

Figure 1. Study flowchart and research questions (RQs). RQ1 addressed participant selection and clinical staging: 103 patients from Shenzhen People’s Hospital were initially screened, with 91 meeting inclusion criteria after exclusions. Participants were classified as Clinical Stage 1–2 (MMSE ≥27; FAQ ≤5) or Clinical Stages 3–6 (MMSE ≤26; FAQ >5), followed by correlation analysis. RQ2 focused on model development: the cohort was randomly divided into training (70%) and testing (30%) sets, with logistic regression analyses performed in the training set to identify predictors for model construction. RQ3 involved internal validation using the testing set to assess model performance. RQ4 concerned external validation: 639 individuals from the ADNI database were initially considered, with 30 matched to the training set via propensity score matching (PSM) based on key predictors. Model performance was evaluated across training, testing, and external validation cohorts using ROC analysis (AUC), decision curve analysis (DCA), and calibration plots.

2.2 Data collection

Demographic, clinical, and laboratory parameters were systematically extracted from electronic medical records. Collected data included age, gender, height, weight, body mass index (BMI), comorbidities (hypertension and hyperlipidemia), systolic and diastolic blood pressure at admission, MMSE scores, CSF biomarkers (pTau181 and Aβ42), and peripheral blood biomarkers such as white blood cell count (WBC), platelet count (PLT), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C).

Based on established evidence, several derived hematological indices were also calculated and included: lymphocyte-to-monocyte ratio (LMR), platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR), prognostic nutritional index (PNI), albumin-to-globulin (A/G) ratio, triglyceride and glucose index (TyG), atherogenic index of plasma (AIP), systemic inflammation response index (SIRI), and systemic immune-inflammation index (SII). These indices have been implicated in nutritional status, metabolic regulation, and immune-inflammatory responses relevant to disease pathophysiology, with several markers previously associated with cognitive impairment (Dobiásová and Frohlich, 2001; Koyama et al., 2016; Cai et al., 2017; Maeda et al., 2019; Hong et al., 2021; Zhou et al., 2021; Dziedzic et al., 2022; Li et al., 2023; Gu et al., 2024; Inoue and Janini Gomes, 2024; Wang et al., 2024; Yang et al., 2024). Detailed formulas for calculating these derived indices are provided in the Supplementary Table S1.

2.3 Model development and evaluation

Prior to statistical analysis, the initial dataset was randomly divided into training and testing sets at a 7:3 ratio. The training set was used for model development, including parameter calibration, variable selection, and algorithmic optimization. The testing set was reserved for internal validation to assess model performance without any parameter adjustments. This approach allowed for evaluation of potential overfitting or underfitting and guided decisions regarding model retraining or alternative methodological approaches.

In the training set, both continuous and binary variables were subjected to univariate logistic regression analysis. The method for dichotomizing continuous variables is detailed in the subsequent section (Douglass, 2020; Demšar and Zupan, 2021). Variables demonstrating statistical significance (p < 0.1) were included in multivariate logistic regression analysis. Predictors retained in the final model were those maintaining significance (p < 0.1) in both univariate and multivariate analyses. Following this stepwise selection process, the identified predictors were incorporated into a multivariate logistic regression model.

A nomogram was developed from this final model to estimate the probability of AD progression to Clinical Stages 3–6. Model performance was evaluated in both the training and testing sets using discrimination and calibration metrics. Calibration curves were generated to compare the nomogram-predicted probabilities with observed outcomes. Discrimination, defined as the model’s ability to differentiate between Clinical Stage 1–2 and Stages 3–6, was assessed using the area under the receiver operating characteristic curve (AUC), with corresponding 95% confidence intervals calculated for both datasets.

The performance of the integrated model was compared against a CSF pTau181-only model. Additional performance metrics—including sensitivity, positive predictive value (PPV), and negative predictive value (NPV)—were calculated using the optimal probability threshold determined by maximizing the Youden index (J = sensitivity + specificity − 1) in the training set.

2.4 External validation using the ADNI cohort

For external validation of the predictive model, data were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database.¹ ADNI was launched in 2003 as a public-private partnership under the leadership of Principal Investigator Michael W. Weiner, MD. The primary objective of ADNI is to determine whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical assessments can effectively track the progression of MCI and early AD.

From the ADNI-1, ADNI-GO, ADNI-2, ADNI-3, and ADNI-4 studies, we initially identified 639 participants with available MMSE scores and serum/CSF biomarker data. To ensure comparability with our internal training set on key predictors, propensity score matching (PSM) was performed based on the variables selected in the final predictive model. This rigorous matching process resulted in a well-balanced external validation cohort. Clinical assessments, including MMSE scores, were conducted, and relevant biomarker analyses were performed at the Biomarker Research Laboratory, University of Pennsylvania, USA.

It should be noted that CSF measurements in the ADNI cohort were conducted using the Luminex xMAP platform (INNO-BIA AlzBio3 kit), whereas those in our internal cohort were performed with the INNOTEST® ELISA platform. Although these assays are known to generate differing absolute values, they have been demonstrated to be highly correlated and comparable in their ability to classify AD pathology (Irwin et al., 2012).

2.5 Statistical analysis

Descriptive statistics were first performed on the original dataset. The normality of all continuous variables was assessed using the Shapiro–Wilk test, and the homogeneity of variance was examined with Levene’s test. Based on these test results, variables were summarized and compared as follows: normally distributed continuous variables were presented as mean ± standard deviation and compared using the independent samples t-test; non-normally distributed continuous variables were expressed as median (first quartile, third quartile) [M (Q₁, Q₃)] and analyzed with the Mann–Whitney U test; categorical variables were summarized as frequency and percentage [n (%)]. A significance level of p < 0.1 was applied for all baseline comparisons. Relationships between variables were examined using Spearman’s correlation analysis, with results visualized in a heatmap.

Univariate logistic regression was conducted for continuous variables, and receiver operating characteristic (ROC) curves were generated. Optimal cutoff values for the PNI index, PLR, and A/G ratio were determined by maximizing the Youden index (J = sensitivity + specificity − 1). Classification thresholds for other continuous variables were established based on previously published studies (Han et al., 2021; Gao et al., 2022; Mehta et al., 2023; Dowling et al., 2024).

3 Results

3.1 Baseline characteristics and univariate analysis

The baseline characteristics of the entire cohort, including both training and testing sets, are summarized in Table 1. In assessing the functional validity of the MMSE-based clinical staging, results demonstrated a significant gradient in functional impairment across the stratified groups, consistent with the underlying staging framework. Participants classified within Clinical Stages 3–6 exhibited markedly worse functional performance on the FAQ compared to those in Clinical Stages 1–2 (median [IQR]: 15 [12–21] vs. 1 [0–3]; Mann–Whitney U = 1194.5, p < 0.001). Comparative distributions of biomarker profiles between Clinical Stage 1–2 and Clinical Stages 3–6 are illustrated in Figure 2, while the results of the Spearman correlation analysis are summarized in Figure 3. Most demographic variables, clinical scores (e.g., MMSE), comorbidities, and biomarker profiles showed no significant differences between the training and testing sets (p > 0.1), supporting their comparability for subsequent model development. A statistically significant difference was observed in the arterial sclerosis index (AIP, p = 0.044), which was therefore excluded from further model development. Optimal cutoff values for the prognostic nutritional index (PNI), platelet-to-lymphocyte ratio (PLR), and albumin-to-globulin (A/G) ratio were determined by maximizing the Youden index. Based on these thresholds, participants were categorized as follows: PLR ≥ 113.22 was classified as the high-PLR group, PNI ≤ 52.05% as the low-PNI group, and A/G ratio ≤1.62 as the low A/G group. Univariate logistic regression analysis identified serum biomarkers significantly associated with advanced clinical stages (Stages 3–6 versus Stages 1–2; p < 0.1). Variables meeting this criterion were entered into an initial multivariable logistic regression model. To ensure model stability and interpretability, variable selection was performed in a stepwise manner: first, the pTau181/Aβ42 ratio was excluded due to high collinearity with CSF pTau181, which attenuated the statistical significance of both terms when included simultaneously. Subsequently, both the A/G ratio ≤1.62 and the PNI ≤ 52.05% lost statistical significance after adjustment and were therefore omitted. This selection process yielded three predictors in the final model: two continuous variables (CSF pTau181 and A/G ratio) and one dichotomized variable (high PLR group). Detailed results of the univariate and multivariable analyses are presented in Table 2.

Table 1

Table 1. Baseline characteristics of the study participants.

Figure 2

Figure 2. Comparison of variable distributions across clinical stages and between datasets. Box plots display the comparative levels of (A) CSF pTau181, (B) A/G ratio, (C) platelet-to-lymphocyte ratio (PLR), and (D) prognostic nutritional index (PNI) between clinical stage 1–2 (coded as 0) and clinical stages 3–6 (coded as 1). Group comparisons were performed using t-test for A/G ratio and Mann–Whitney U test for all other indicators. Statistically significant p-values are annotated.

Figure 3

Figure 3. Spearman correlation heatmap of clinical stage and different variables. Spearman correlation heatmap depicting relationships between clinical stages, biomarkers, and metabolic indices, including MMSE scores, pTau181/Aβ42 ratio, inflammatory markers, and metabolic parameters. Color gradient represents correlation strength, with dark blue indicating strong positive correlations and dark red indicating strong negative correlations.

Table 2

Table 2. The results of univariate and multivariate logistic regression.

3.2 Predictive model and performance

The integrated model, developed from multivariate logistic regression in the training set, demonstrated excellent discrimination for identifying Clinical Stages 3–6, with an area under the receiver operating characteristic curve (AUC) of 0.92 (95% CI: 0.83–1.00; Figure 4B). The model maintained robust performance in the independent testing set, achieving an AUC of 0.86 (95% CI: 0.70–1.00; Figure 4C). Notably, the combined model showed significantly improved discrimination compared against a CSF pTau181-only model (AUC = 0.78, 95% CI: 0.64–0.92; Figures 4E-G). Additional performance metrics of the final model on the testing set are listed in Table 3.

Figure 4

Figure 4. Performance evaluation of the integrated model and its individual components. (A) Nomogram of the final integrated model for discriminating between AD clinical stages 3–6 and stage 2. Receiver operating characteristic (ROC) curves demonstrate the discriminatory performance of (B) the integrated model in the training set (AUC = 0.92, 95% CI: 0.83–1.00), (C) the integrated model in the testing set (AUC = 0.86, 95% CI: 0.70–1.00), and (D) the external validation set (AUC = 0.83, 95% CI: 0.67–0.98). Comparative ROC analyses show the performance of (E) CSF pTau181 alone (AUC = 0.78, 95% CI: 0.64–0.92), (F) the albumin-to-globulin (A/G) ratio (AUC = 0.76, 95% CI: 0.52–1.00), and (G) the platelet-to-lymphocyte ratio (PLR ≥ 113.22; AUC = 0.69, 95% CI: 0.54–0.84).

Table 3

Table 3. Performance of the integrated predictive model in the testing set.

A nomogram was constructed based on the final multivariate regression model to visually estimate the probability of an AD patient belonging to Clinical Stages 3–6 (Figure 4A). Calibration performance in the testing cohort showed strong agreement between predicted probabilities and observed outcomes. The bias-corrected curve, derived from 1,000 bootstrap resamples, closely followed the ideal calibration line, with a mean absolute error (MAE) of 0.039, indicating minimal deviation between predictions and observations (Figures 5A,B). Decision curve analysis (DCA) revealed that the model provided higher net benefit than the “treat all” strategy across a wide range of threshold probabilities, although within a narrow interval (approximately 25–50%), the net benefit was marginally lower than that of the “treat all” approach (Figures 5D,E).

Figure 5

Figure 5. Calibration and clinical utility of the integrated predictive model. Calibration curves for the integrated model in the (A) training set, (B) testing set, and (C) external validation set. The dashed 45-degree line represents ideal calibration, while the solid bias-corrected line indicates model performance after overfitting adjustment. Closer alignment with the diagonal indicates better agreement between predicted probabilities and observed outcomes. Decision curve analysis (DCA) evaluating the clinical utility of the integrated model in the (D) training set, (E) testing set, and (F) external validation set. The DCA curves demonstrate the net benefit of the model across different threshold probabilities.

3.3 External validation with propensity score matching

External validation was conducted using data from the ADNI cohort, which initially included 639 individuals. Propensity score matching (PSM) was applied to align the ADNI samples with the training set based on three key predictors in the final model: CSF pTau181, A/G ratio, and high PLR status (PLR ≥ 113.22). After matching, 30 individuals from the ADNI cohort were successfully paired with training set participants, achieving balanced distributions as indicated by minimal standardized mean differences (SMDs; Figure 6; Table 4).

Figure 6

Figure 6. Distribution balance for distance after propensity score matching. Density plots illustrating the distribution of propensity scores between the internal training set (labeled as 1) and the ADNI external validation cohort (labeled as 0). The left panel shows the initial imbalance in the unmatched sample, while the right panel demonstrates substantially improved balance achieved after propensity score matching.

Table 4

Table 4. Propensity score matching results for the external validation set.

In the external validation set, the model exhibited robust discriminatory power with an AUC of 0.83 (95% CI: 0.67–0.98) despite the inter-platform assay differences (Figure 4D). Moreover, its clinical utility and calibration were consistently affirmed by decision curve and calibration curve analyses within the same cohort (Figures 5C,F).

4 Discussion

In this study, we developed and validated a predictive model for AD clinical progression by integrating CSF pTau181 with serum biomarkers—the albumin-to-globulin (A/G) ratio and the platelet-to-lymphocyte ratio (PLR). The model demonstrated robust discrimination and good calibration across both internal and external validation cohorts, highlighting the contributions of nutritional status and systemic inflammation to the AD clinical continuum.

4.1 Nutritional status and the A/G ratio

Our findings indicate that a higher A/G ratio is associated with a reduced risk of AD progression, suggesting a potential protective role. The A/G ratio reflects the balance between protein nutrition, represented by albumin, and immune activity, represented by globulin. An elevated A/G ratio typically signifies adequate nutritional reserves and antioxidant capacity, whereas a lower ratio may indicate chronic inflammation or malnutrition. This balance may be modulated by the brain-gut axis, through which gut health influences systemic inflammation and nutrient absorption—key determinants of albumin and globulin levels (Zhou et al., 2022).

Previous epidemiological studies have consistently linked a lower A/G ratio to an increased risk of cognitive impairment (Koyama et al., 2016; Maeda et al., 2019). Recent analyses of NHANES data further support a non-linear relationship between the A/G ratio and cognitive function in older adults (Yang et al., 2024). Our results support and extend these observations, reinforcing the notion that nutritional status serves as a modifiable protective factor against AD progression.

Mechanistically, albumin not only functions as a transport protein but also exhibits antioxidant and anti-inflammatory properties, enabling it to scavenge free radicals and bind amyloid proteins. The promising results from the AMBAR clinical trial, which demonstrated benefits of albumin-based plasma exchange in AD patients, provide direct human evidence supporting the neuroprotective and anti-Aβ roles of albumin, potentially mediated through stabilization of the blood–brain barrier (Boada et al., 2020). Furthermore, malnutrition can impair neuronal resilience by limiting substrates essential for synaptic maintenance. Experimental studies have shown that protein-energy deficiency downregulates the expression of key synaptic proteins (e.g., BDNF, GAP-43) and induces detrimental hippocampal ultrastructural changes, deficits that can be partially reversed by dietary supplementation (Prosser-Loose et al., 2010; Sato et al., 2021).

Despite the strong biological plausibility, some caution is warranted when extrapolating these findings beyond the study population. The primary cohort was recruited from a single hospital in China, and systemic inflammatory and nutritional biomarkers may vary across populations owing to differences in ethnicity, dietary patterns, and healthcare context. Although external validation was conducted using the ADNI cohort, this analysis was retrospective and based on a limited subset of the overall ADNI population, which may constrain direct translation to real-world clinical settings. Thus, while the observed biological associations appear robust, the predictive performance of the model requires further evaluation in diverse, prospective, multicenter cohorts.

Clinically, these findings underscore the importance of nutritional interventions in AD management. Substantial evidence has demonstrated that dietary patterns such as the Mediterranean diet—rich in vegetables, fruits, whole grains, and fish—significantly reduce the risk of cognitive decline and AD (Scarmeas et al., 2018). Our study provides a mechanistic rationale for such interventions, suggesting that maintaining an optimal A/G ratio through adequate nutrition may represent a viable strategy to slow disease progression.

4.2 Systemic inflammation and the PLR

Conversely, we found that an elevated PLR was associated with a higher risk of AD progression. The PLR reflects the balance between pro-inflammatory activity (represented by platelets) and immune regulation (represented by lymphocytes). A high PLR suggests a state of enhanced systemic inflammation coupled with compromised immune protection.

Systemic inflammation is postulated to exacerbate AD pathology through peripheral-central immune crosstalk. This process may be further amplified by the disruption of brain energy metabolism, as systemic inflammation can impair cerebral glucose utilization and mitochondrial function, thereby aggravating neuronal vulnerability and pathological progression (Yu et al., 2022). Potential mechanisms include: (1) systemic inflammation activating inflammatory pathways (e.g., the GSDMD signaling) in cerebral endothelial cells, leading to blood–brain barrier damage (Wei et al., 2024); (2) peripheral inflammatory mediators infiltrating the central nervous system, activating microglia and triggering neuroinflammation and tau hyperphosphorylation (Heneka et al., 2015; Leng and Edison, 2021); (3) disruption of immune homeostasis impairing intrinsic neuroprotective mechanisms. Our results align with previous studies linking elevated peripheral inflammatory markers, including PLR and NLR, to cognitive decline, brain atrophy, and Aβ deposition (Han et al., 2021; Li et al., 2023; Zheng et al., 2025). Therefore, in addition to nutritional support, anti-inflammatory strategies—such as dietary modifications rich in polyunsaturated fatty acids and polyphenols, and early management of comorbid inflammatory conditions—should be emphasized in AD care.

4.3 Model integration and clinical implications

The integration of CSF pTau181, a core biomarker of AD neurofibrillary pathology, with peripheral indicators of nutrition (A/G ratio) and inflammation (PLR), provides a more holistic view of the disease. This combined model suggests that AD clinical progression is driven not only by central tau pathology but also by interplay with systemic physiological states. This multi-system perspective aligns with emerging research and could enhance risk stratification in clinical practice, potentially identifying patients who might benefit from integrated management strategies targeting both central pathology and peripheral modifiers.

4.4 Limitations

Several limitations of this study should be acknowledged. First, the relatively small sample size of our primary cohort increases the risk of model overfitting, and the wide confidence intervals observed in the testing set warrant cautious interpretation. In addition, although external validation was conducted using the ADNI dataset, the matched validation cohort constitutes only a small subset of the overall ADNI population. A further methodological consideration pertains to the absence of a formal prospective sample size calculation. While all eligible participants from available retrospective datasets were included, the resulting cohort size remains below the thresholds typically recommended in contemporary guidelines for fully optimized predictive modeling. Given the limited dataset, the dichotomization of selected continuous variables represents a pragmatic trade-off to enable construction of a bedside-friendly nomogram, while more complex machine-learning approaches are not suitable for small-sample modeling. Consequently, the current model should be interpreted as a preliminary proof of concept. External validation in larger, independent cohorts will be essential to confirm the generalizability of the findings (Riley et al., 2020; Pate et al., 2023; Efthimiou et al., 2024). Second, the primary aim of the nomogram is to facilitate risk stratification, not to direct individualized treatment. In line with this function, we observed a decline in net benefit across intermediate predicted probability ranges (25–50%). Thus, the model may offer limited clinical utility for patients within this interval, thereby indicating its greater applicability at the lower and upper extremes of the risk spectrum. Third, although Aβ pathology is a cornerstone of AD, Aβ markers were not retained in our final multivariate model, likely due to statistical reasons rather than biological irrelevance (Janelidze et al., 2016). Previous studies have underscored the high diagnostic value of the pTau181/Aβ42 ratio in both CSF and serum (Ortner et al., 2022; Lehmann et al., 2025), and our Spearman correlation analysis also confirmed a significant association between this ratio and AD clinical stages. Nonetheless, it did not retain statistical significance in the final multivariate logistic regression, and therefore was excluded from model construction. Certain factors, such as APOE genotype, vascular comorbidities, and neuroimaging biomarkers, were intentionally omitted from the model. This strategic omission represents a pragmatic design choice to ensure clinical feasibility and scalability by relying on more readily obtainable CSF and routine blood-based measures.

Fourth, Phosphorylated tau at threonine 181 (pTau181) is a core CSF biomarker of AD–related neurofibrillary pathology. Its positive association with disease presence and severity has been consistently demonstrated across diverse populations and analytical platforms, underscoring its biological robustness and generalizability (Janelidze et al., 2020).

The Luminex xMAP platform was among the first high-throughput immunoassay systems adopted for quantifying CSF tau and pTau181. Its clinical relevance to AD severity and neuropathological burden was established in large multicenter cohorts, including the Alzheimer’s Disease Neuroimaging Initiative (ADNI), during the mid- to late-2000s (Olsson et al., 2005; Shaw et al., 2009). Subsequently, enzyme-linked immunosorbent assay (ELISA)–based methods—valued for their technical simplicity, commercial availability, and high analytical stability—have gained widespread use in both research and clinical laboratory practice. These assays have likewise demonstrated strong correlations between CSF pTau181 levels and AD severity (Blennow et al., 2010; Janelidze et al., 2020).

Notably, methodological comparisons have shown that although ELISA- and xMAP-based assays yield systematically different absolute concentrations, measurements from the two platforms are highly correlated and exhibit comparable diagnostic performance for AD pathology (Irwin et al., 2012; Lleó et al., 2018). These observations indicate that both analytical approaches reliably capture the same underlying biological process—namely, the progressive increase in pTau181 along the AD continuum.

In the present study, despite differences in the CSF pTau181 assay platform between the internal and external (ADNI) cohorts, the predictive model retained good discriminative performance in the external validation set (area under the curve [AUC] = 0.83, 95% confidence interval [CI]: 0.67–0.98; Figure 4D). Consistent calibration and favorable decision-curve analyses further supported the model’s robustness in this independent population (Figures 5C,F). Accordingly, the primary objective of the external validation was not to compare absolute biomarker values across platforms, but to evaluate whether the predictive relationship identified by the model—which integrates CSF pTau181 with the serum albumin-to-globulin ratio and platelet-to-lymphocyte ratio—could be generalized to an independently ascertained cohort. The maintained performance in the ADNI cohort suggests that this integrative biological signal remains informative even when the central biomarker component is quantified using a different, yet validated, analytical platform. It should be noted, however, that inter-platform differences may influence absolute risk estimation; therefore, platform-specific recalibration would be necessary before clinical translation.

5 Conclusion

Despite these limitations, our integrated model marks meaningful advance toward personalized and accessible AD prognosis. It establishes a practical and accessible framework that assists clinicians in identifying AD patients who have progressed to Clinical Stages 3–6, thereby facilitating more accurate prognostic evaluation and tailored treatment planning. It should be noted that the present model is designed to identify the transition to the stage of objective cognitive impairment. The separate and distinct question of predicting the rate or likelihood of progression from mild cognitive impairment (MCI) to dementia warrants future investigation through dedicated longitudinal study designs. Furthermore, our findings contribute to a deeper understanding of the role played by nutritional status and immune response in the pathophysiological progression of AD, providing a basis for future research on integrated, multi-dimensional approaches to AD care and intervention.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics statement

The studies involving humans were approved by the Ethics Committee of Shenzhen People’s Hospital. The studies were conducted in accordance with the local legislation and institutional requirements. The ethics committee/institutional review board waived the requirement of written informed consent for participation from the participants or the participants’ legal guardians/next of kin because It is a retrospective analysis. All ADNI sites acquired local institutional review board approval and informed subject consent for participation in the ADNI study.

Group members of Alzheimer’s Disease Neuroimaging Initiative

Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf.

Author contributions

XW: Writing – original draft, Writing – review & editing, Data curation, Formal analysis, Investigation, Methodology, Software. YT: Methodology, Writing – original draft. QiwZ: Writing – original draft, Investigation, Writing – review & editing. BX: Writing – original draft, Writing – review & editing, Formal analysis, Methodology, Visualization. SZ: Writing – original draft, Writing – review & editing, Data curation, Investigation. QiaZ: Data curation, Investigation, Writing – original draft, Writing – review & editing. MG: Investigation, Writing – original draft, Writing – review & editing, Methodology. LZ: Writing – original draft, Writing – review & editing, Conceptualization, Funding acquisition, Project administration, Resources, Supervision.

Funding

The author(s) declared that financial support was received for this work and/or its publication. The study was registered in the Chinese Medical Research Registration and Filing System (Approval No. MR-44-24-047576). This work was supported by the Guangdong Provincial Key S&T Program (2018B030336001).

Acknowledgments

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

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 of this manuscript.

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/fnagi.2026.1728675/full#supplementary-material

Footnotes

1. ^adni.loni.usc.edu

References

Balachandran, V. P., Gonen, M., Smith, J. J., and DeMatteo, R. P. (2015). Nomograms in oncology: more than meets the eye. Lancet Oncol. 16, e173–e180. doi: 10.1016/s1470-2045(14)71116-7

Crossref Full Text | Google Scholar

Benedet, A. L., Milà-Alomà, M., Vrillon, A., Ashton, N. J., Pascoal, T. A., Lussier, F., et al. (2021). Differences between plasma and cerebrospinal fluid glial fibrillary acidic protein levels across the Alzheimer disease continuum. JAMA Neurol. 78, 1471–1483. doi: 10.1001/jamaneurol.2021.3671

Crossref Full Text | Google Scholar

Blennow, K., Hampel, H., Weiner, M., and Zetterberg, H. (2010). Cerebrospinal fluid and plasma biomarkers in Alzheimer disease. Nat. Rev. Neurol. 6, 131–144. doi: 10.1038/nrneurol.2010.4

Crossref Full Text | Google Scholar

Boada, M., López, O. L., Olazarán, J., Núñez, L., Pfeffer, M., Paricio, M., et al. (2020). A randomized, controlled clinical trial of plasma exchange with albumin replacement for Alzheimer's disease: primary results of the AMBAR study. Alzheimers Dement. 16, 1412–1425. doi: 10.1002/alz.12137

Crossref Full Text | Google Scholar

Cai, G., Shi, G., Xue, S., and Lu, W. (2017). The atherogenic index of plasma is a strong and independent predictor for coronary artery disease in the Chinese Han population. Medicine (Baltimore) 96:e8058. doi: 10.1097/md.0000000000008058

Crossref Full Text | Google Scholar

Chen, S., Ou, R., Wei, Q., Fu, J., Zhao, B., Chen, X., et al. (2025). Identification of risk factors and development of a predictive nomogram for sarcopenia in Alzheimer's disease. Alzheimers Dement. 21:e14503. doi: 10.1002/alz.14503

Crossref Full Text | Google Scholar

Cohen, S., van Dyck, C. H., Gee, M., Doherty, T., Kanekiyo, M., Dhadda, S., et al. (2023). Lecanemab Clarity AD: quality-of-life results from a randomized, double-blind phase 3 trial in early Alzheimer's disease. J. Prev Alzheimers Dis. 10, 771–777. doi: 10.14283/jpad.2023.123

Crossref Full Text | Google Scholar

Demšar, J., and Zupan, B. (2021). Hands-on training about overfitting. PLoS Comput. Biol. 17:e1008671. doi: 10.1371/journal.pcbi.1008671

Crossref Full Text | Google Scholar

Dobiásová, M., and Frohlich, J. (2001). The plasma parameter log (TG/HDL-C) as an atherogenic index: correlation with lipoprotein particle size and esterification rate in apoB-lipoprotein-depleted plasma (FER(HDL)). Clin. Biochem. 34, 583–588. doi: 10.1016/s0009-9120(01)00263-6

Crossref Full Text | Google Scholar

Douglass, M. J. J. (2020). Book review: hands-on machine learning with Scikit-learn, Keras, and Tensorflow, 2nd edition by Aurélien Géron. Phys. Eng. Sci. Med. 43, 1135–1136. doi: 10.1007/s13246-020-00913-z

Crossref Full Text | Google Scholar

Dowling, G. P., Daly, G. R., Hegarty, A., Hembrecht, S., Bracken, A., Toomey, S., et al. (2024). Predictive value of pretreatment circulating inflammatory response markers in the neoadjuvant treatment of breast cancer: meta-analysis. Br. J. Surg. 111:znae132. doi: 10.1093/bjs/znae132

Crossref Full Text | Google Scholar

Dziedzic, E. A., Gąsior, J. S., Tuzimek, A., Paleczny, J., Junka, A., Dąbrowski, M., et al. (2022). Investigation of the associations of novel inflammatory biomarkers-systemic inflammatory index (SII) and systemic inflammatory response index (SIRI)-with the severity of coronary artery disease and acute coronary syndrome occurrence. Int. J. Mol. Sci. 23:9533. doi: 10.3390/ijms23179553

Crossref Full Text | Google Scholar

Efthimiou, O., Seo, M., Chalkou, K., Debray, T., Egger, M., and Salanti, G. (2024). Developing clinical prediction models: a step-by-step guide. BMJ 386:e078276. doi: 10.1136/bmj-2023-078276

Crossref Full Text | Google Scholar

Gao, S., Tang, W., Zuo, B., Mulvihill, L., Yu, J., and Yu, Y. (2022). The predictive value of neutrophil-to-lymphocyte ratio for overall survival and pathological complete response in breast cancer patients receiving neoadjuvant chemotherapy. Front. Oncol. 12:1065606. doi: 10.3389/fonc.2022.1065606

Crossref Full Text | Google Scholar

Graff-Radford, J., Yong, K. X. X., Apostolova, L. G., Bouwman, F. H., Carrillo, M., Dickerson, B. C., et al. (2021). New insights into atypical Alzheimer's disease in the era of biomarkers. Lancet Neurol. 20, 222–234. doi: 10.1016/s1474-4422(20)30440-3

Crossref Full Text | Google Scholar

Gu, L., Xia, Z., Qing, B., Wang, W., Chen, H., Wang, J., et al. (2024). Systemic inflammatory response index (SIRI) is associated with all-cause mortality and cardiovascular mortality in population with chronic kidney disease: evidence from NHANES (2001-2018). Front. Immunol. 15:1338025. doi: 10.3389/fimmu.2024.1338025

Crossref Full Text | Google Scholar

Guzman-Martinez, L., Calfío, C., Farias, G. A., Vilches, C., Prieto, R., and Maccioni, R. B. (2021). New Frontiers in the prevention, diagnosis, and treatment of Alzheimer's disease. J Alzheimer's Dis 82, S51–s63. doi: 10.3233/jad-201059

Crossref Full Text | Google Scholar

Han, X., Liu, S., Yang, G., Hosseinifard, H., Imani, S., Yang, L., et al. (2021). Prognostic value of systemic hemato-immunological indices in uterine cervical cancer: a systemic review, meta-analysis, and meta-regression of observational studies. Gynecol. Oncol. 160, 351–360. doi: 10.1016/j.ygyno.2020.10.011

Crossref Full Text | Google Scholar

Heneka, M. T., Carson, M. J., El Khoury, J., Landreth, G. E., Brosseron, F., Feinstein, D. L., et al. (2015). Neuroinflammation in Alzheimer's disease. Lancet Neurol. 14, 388–405. doi: 10.1016/s1474-4422(15)70016-5

Crossref Full Text | Google Scholar

Hong, S., Han, K., and Park, C. Y. (2021). The insulin resistance by triglyceride glucose index and risk for dementia: population-based study. Alzheimer's Res Ther 13:9. doi: 10.1186/s13195-020-00758-4

Crossref Full Text | Google Scholar

Inoue, D. S., and Janini Gomes, M. (2024). Integrative insights into PNI: low-grade chronic inflammation, skeletal muscle wasting, and brain impairments. Brain Behav Immun Health 40:100838. doi: 10.1016/j.bbih.2024.100838

Crossref Full Text | Google Scholar

Irwin, D. J., McMillan, C. T., Toledo, J. B., Arnold, S. E., Shaw, L. M., Wang, L. S., et al. (2012). Comparison of cerebrospinal fluid levels of tau and aβ 1-42 in Alzheimer disease and frontotemporal degeneration using 2 analytical platforms. Arch. Neurol. 69, 1018–1025. doi: 10.1001/archneurol.2012.26

Crossref Full Text | Google Scholar

Jack, C. R. Jr., Andrews, J. S., Beach, T. G., Buracchio, T., Dunn, B., Graf, A., et al. (2024). Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association workgroup. Alzheimers Dement. 20, 5143–5169. doi: 10.1002/alz.13859

Crossref Full Text | Google Scholar

Jack, C. R. Jr., Bennett, D. A., Blennow, K., Carrillo, M. C., Dunn, B., Haeberlein, S. B., et al. (2018). NIA-AA research framework: toward a biological definition of Alzheimer's disease. Alzheimers Dement. 14, 535–562. doi: 10.1016/j.jalz.2018.02.018

Crossref Full Text | Google Scholar

Janelidze, S., Mattsson, N., Palmqvist, S., Smith, R., Beach, T. G., Serrano, G. E., et al. (2020). Plasma P-tau181 in Alzheimer's disease: relationship to other biomarkers, differential diagnosis, neuropathology and longitudinal progression to Alzheimer's dementia. Nat. Med. 26, 379–386. doi: 10.1038/s41591-020-0755-1

Crossref Full Text | Google Scholar

Janelidze, S., Stomrud, E., Palmqvist, S., Zetterberg, H., van Westen, D., Jeromin, A., et al. (2016). Plasma β-amyloid in Alzheimer's disease and vascular disease. Sci. Rep. 6:26801. doi: 10.1038/srep26801

Crossref Full Text | Google Scholar

Koyama, T., Kuriyama, N., Ozaki, E., Matsui, D., Watanabe, I., Miyatani, F., et al. (2016). Serum albumin to globulin ratio is related to cognitive decline via reflection of homeostasis: a nested case-control study. BMC Neurol. 16:253. doi: 10.1186/s12883-016-0776-z

Crossref Full Text | Google Scholar

Koyama, A., O'Brien, J., Weuve, J., Blacker, D., Metti, A. L., and Yaffe, K. (2013). The role of peripheral inflammatory markers in dementia and Alzheimer's disease: a meta-analysis. J. Gerontol. A Biol. Sci. Med. Sci. 68, 433–440. doi: 10.1093/gerona/gls187

Crossref Full Text | Google Scholar

Lehmann, S., Gabelle, A., Duchiron, M., Busto, G., Morchikh, M., Delaby, C., et al. (2025). Comparative performance of plasma pTau181/Aβ42, pTau217/Aβ42 ratios, and individual measurements in detecting brain amyloidosis. EBioMedicine 117:105805. doi: 10.1016/j.ebiom.2025.105805

Crossref Full Text | Google Scholar

Leng, F., and Edison, P. (2021). Neuroinflammation and microglial activation in Alzheimer disease: where do we go from here? Nat. Rev. Neurol. 17, 157–172. doi: 10.1038/s41582-020-00435-y

Crossref Full Text | Google Scholar

Li, J. Q., Zhang, Y. R., Wang, H. F., Guo, Y., Shen, X. N., Li, M. M., et al. (2023). Exploring the links among peripheral immunity, biomarkers, cognition, and neuroimaging in Alzheimer's disease. Alzheimers Dement (Amst) 15:e12517. doi: 10.1002/dad2.12517

Crossref Full Text | Google Scholar

Lleó, A., Irwin, D. J., Illán-Gala, I., McMillan, C. T., Wolk, D. A., Lee, E. B., et al. (2018). A 2-step cerebrospinal algorithm for the selection of frontotemporal lobar degeneration subtypes. JAMA Neurol. 75, 738–745. doi: 10.1001/jamaneurol.2018.0118

Crossref Full Text | Google Scholar

Lv, X., Cheng, Z., Wang, Q., Gao, F., Dai, L., Du, C., et al. (2023). High burdens of phosphorylated tau protein and distinct precuneus atrophy in sporadic early-onset Alzheimer's disease. Sci Bull (Beijing) 68, 2817–2826. doi: 10.1016/j.scib.2023.10.019

Crossref Full Text | Google Scholar

Lv, J., Ren, H., Guo, X., Meng, C., Fei, J., Mei, H., et al. (2022). Nomogram predicting bullying victimization in adolescents. J. Affect. Disord. 303, 264–272. doi: 10.1016/j.jad.2022.02.037

Crossref Full Text | Google Scholar

Maeda, S., Takeya, Y., Oguro, R., Akasaka, H., Ryuno, H., Kabayama, M., et al. (2019). Serum albumin/globulin ratio is associated with cognitive function in community-dwelling older people: the septuagenarians, octogenarians, nonagenarians investigation with centenarians study. Geriatr Gerontol Int 19, 967–971. doi: 10.1111/ggi.13751

Crossref Full Text | Google Scholar

Marshall, G. A., Zoller, A. S., Lorius, N., Amariglio, R. E., Locascio, J. J., Johnson, K. A., et al. (2015). Functional activities questionnaire items that best discriminate and predict progression from clinically Normal to mild cognitive impairment. Curr. Alzheimer Res. 12, 493–502. doi: 10.2174/156720501205150526115003

Crossref Full Text | Google Scholar

McDade, E. M. (2022). Alzheimer Disease. Continuum 28, 648–675. doi: 10.1212/CON.0000000000001131

Crossref Full Text | Google Scholar

Mehta, N. H., Zhou, L., Li, Y., McIntire, L. B., Nordvig, A., Butler, T., et al. (2023). Peripheral immune cell imbalance is associated with cortical beta-amyloid deposition and longitudinal cognitive decline. Sci. Rep. 13:8847. doi: 10.1038/s41598-023-34012-2

Crossref Full Text | Google Scholar

Olsson, A., Vanderstichele, H., Andreasen, N., De Meyer, G., Wallin, A., Holmberg, B., et al. (2005). Simultaneous measurement of beta-amyloid(1-42), total tau, and phosphorylated tau (Thr181) in cerebrospinal fluid by the xMAP technology. Clin. Chem. 51, 336–345. doi: 10.1373/clinchem.2004.039347

Crossref Full Text | Google Scholar

Ortner, M., Goldhardt, O., Diehl-Schmid, J., Yakushev, I., Lanz, K., Hedderich, D. M., et al. (2022). Elecsys cerebrospinal fluid assays accurately distinguish Alzheimer's disease from frontotemporal lobar degeneration. J. Prev Alzheimers Dis. 9, 491–498. doi: 10.14283/jpad.2022.27

Crossref Full Text | Google Scholar

Pate, A., Riley, R. D., Collins, G. S., van Smeden, M., Van Calster, B., Ensor, J., et al. (2023). Minimum sample size for developing a multivariable prediction model using multinomial logistic regression. Stat. Methods Med. Res. 32, 555–571. doi: 10.1177/09622802231151220

Crossref Full Text | Google Scholar

Pereira, J. B., Janelidze, S., Smith, R., Mattsson-Carlgren, N., Palmqvist, S., Teunissen, C. E., et al. (2021). Plasma GFAP is an early marker of amyloid-β but not tau pathology in Alzheimer's disease. Brain 144, 3505–3516. doi: 10.1093/brain/awab223

Crossref Full Text | Google Scholar

Petersen, R. C., Caracciolo, B., Brayne, C., Gauthier, S., Jelic, V., and Fratiglioni, L. (2014). Mild cognitive impairment: a concept in evolution. J. Intern. Med. 275, 214–228. doi: 10.1111/joim.12190

Crossref Full Text | Google Scholar

Prosser-Loose, E. J., Verge, V. M., Cayabyab, F. S., and Paterson, P. G. (2010). Protein-energy malnutrition alters hippocampal plasticity-associated protein expression following global ischemia in the gerbil. Curr. Neurovasc. Res. 7, 341–360. doi: 10.2174/156720210793180792

Crossref Full Text | Google Scholar

Riley, R. D., Ensor, J., Snell, K. I. E., Harrell, F. E. Jr., Martin, G. P., Reitsma, J. B., et al. (2020). Calculating the sample size required for developing a clinical prediction model. BMJ 368:m441. doi: 10.1136/bmj.m441

Crossref Full Text | Google Scholar

Sato, H., Takado, Y., Toyoda, S., Tsukamoto-Yasui, M., Minatohara, K., Takuwa, H., et al. (2021). Neurodegenerative processes accelerated by protein malnutrition and decelerated by essential amino acids in a tauopathy mouse model. Sci. Adv. 7:eabd5046. doi: 10.1126/sciadv.abd5046

Crossref Full Text | Google Scholar

Scarmeas, N., Anastasiou, C. A., and Yannakoulia, M. (2018). Nutrition and prevention of cognitive impairment. Lancet Neurol. 17, 1006–1015. doi: 10.1016/s1474-4422(18)30338-7

Crossref Full Text | Google Scholar

Selma-Gonzalez, J., Rubio-Guerra, S., García-Castro, J., Vera-Campuzano, E., Sala, I., Sánchez-Saudinós, M. B., et al. (2025). Association of Plasma Phosphorylated tau 217 with clinical deterioration across Alzheimer disease stages. Neurology 105:e213769. doi: 10.1212/wnl.0000000000213769

Crossref Full Text | Google Scholar

Shaw, L. M., Vanderstichele, H., Knapik-Czajka, M., Clark, C. M., Aisen, P. S., Petersen, R. C., et al. (2009). Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects. Ann. Neurol. 65, 403–413. doi: 10.1002/ana.21610

Crossref Full Text | Google Scholar

Tahami Monfared, A. A., Tafazzoli, A., Ye, W., Chavan, A., and Zhang, Q. (2022). Long-term health outcomes of Lecanemab in patients with early Alzheimer's disease using simulation modeling. Neurol Ther 11, 863–880. doi: 10.1007/s40120-022-00350-y

Crossref Full Text | Google Scholar

Tahami Monfared, A. A., Ye, W., Sardesai, A., Folse, H., Chavan, A., Aruffo, E., et al. (2023). A path to improved Alzheimer's care: simulating long-term health outcomes of Lecanemab in early Alzheimer's disease from the CLARITY AD trial. Neurol Ther 12, 863–881. doi: 10.1007/s40120-023-00473-w

Crossref Full Text | Google Scholar

Tatulian, S. A. (2022). Challenges and hopes for Alzheimer's disease. Drug Discov. Today 27, 1027–1043. doi: 10.1016/j.drudis.2022.01.016

Crossref Full Text | Google Scholar

van der Lee, S. J., Teunissen, C. E., Pool, R., Shipley, M. J., Teumer, A., Chouraki, V., et al. (2018). Circulating metabolites and general cognitive ability and dementia: evidence from 11 cohort studies. Alzheimers Dement. 14, 707–722. doi: 10.1016/j.jalz.2017.11.012

Crossref Full Text | Google Scholar

van Oostveen, W. M., and de Lange, E. C. M. (2021). Imaging techniques in Alzheimer's disease: a review of applications in early diagnosis and longitudinal monitoring. Int. J. Mol. Sci. 22:2110. doi: 10.3390/ijms22042110

Crossref Full Text | Google Scholar

Wang, Y., Zhang, G., Shen, Y., Zhao, P., Sun, H., Ji, Y., et al. (2024). Relationship between prognostic nutritional index and post-stroke cognitive impairment. Nutr. Neurosci. 27, 1330–1340. doi: 10.1080/1028415x.2024.2330786

Crossref Full Text | Google Scholar

Wei, C., Jiang, W., Wang, R., Zhong, H., He, H., Gao, X., et al. (2024). Brain endothelial GSDMD activation mediates inflammatory BBB breakdown. Nature 629, 893–900. doi: 10.1038/s41586-024-07314-2

Crossref Full Text | Google Scholar

Yang, H., Liao, Z., Zhou, Y., Gao, Z., and Mao, Y. (2024). Non-linear relationship of serum albumin-to-globulin ratio and cognitive function in American older people: a cross-sectional national health and nutrition examination survey 2011-2014 (NHANES) study. Front. Public Health 12:1375379. doi: 10.3389/fpubh.2024.1375379

Crossref Full Text | Google Scholar

Yu, L., Jin, J., Xu, Y., and Zhu, X. (2022). Aberrant energy metabolism in Alzheimer's disease. J Transl Int Med 10, 197–206. doi: 10.2478/jtim-2022-0024

Crossref Full Text | Google Scholar

Zheng, Y., Yu, Y., Gao, L., Yu, M., Jiang, L., and Zhu, Q. (2025). Association of red blood cell count, hemoglobin concentration, and inflammatory indices with cognitive impairment severity in Alzheimer's disease. Sci. Rep. 15:17425. doi: 10.1038/s41598-025-02468-z

Crossref Full Text | Google Scholar

Zhou, J., Ma, L., Zhao, L., Sheng, J., Xu, Y., Chen, J., et al. (2021). Association between the prognostic nutritional index and cognitive function among older adults in the United States: a population-based study. J Alzheimer's Dis 83, 819–831. doi: 10.3233/jad-210141

Crossref Full Text | Google Scholar

Zhou, G., Yin, Y., Huan, X., Zhuang, Y., Xu, S., Liu, J., et al. (2022). The potential role of the brain-gut Axis in the development and progression of Alzheimer's disease. J Transl Int Med 10, 89–91. doi: 10.2478/jtim-2022-0016

Crossref Full Text | Google Scholar