Skip to main content
NCBI home page
JAMA Network logoLink to JAMA Network
. 2026 Jul 28;9(7):e2625730. doi: 10.1001/jamanetworkopen.2026.25730

Plasma Amyloid-β 1-40 and Postoperative Delirium After Cardiac Surgery

Christine A F von Arnim 1,2,✉, Monika Sadlonova 1,2,3,4,5, Carlotta Derad 6, Maike Hohberg 6, Christopher M Celano 5,7, Irina Günther 8, Tobias Titsch 8, Iryna Krasiuk 8, Katharina Packroß 8, Matilda-Marie Becker 8, Alina Isabel Rediske 8, Nicholas Paul Süttmann 8, Clara Maria Knopp 1, Stephanie Heinemann 1, Paul T Itting 2,3, Ingo Kutschka 9, Barbara Morgado 8,10, Hermann Esselmann 8, Hassina Baraki 9, Thomas Asendorf 6, Niels Hansen 8, Jens Wiltfang 8,10,11, for the FIND Delirium Risk Factors (FINDERI) investigators
PMCID: PMC13416913  PMID: 42518233

Key Points

Question

Are plasma Alzheimer disease biomarkers associated with postoperative delirium (POD) risk in older adults undergoing elective cardiac surgery?

Findings

In this cohort study of 491 patients, elevated plasma amyloid-β 1-40 (Aβ1-40) levels were significantly associated with higher POD risk, independent of age and baseline cognition. Multivariable models combining Aβ1-40 with clinical factors and Montreal Cognitive Assessment showed improved risk estimation of POD compared with clinical variables alone.

Meaning

These findings suggest that preoperative plasma Aβ1-40 may serve as a modest but independent risk factor for POD in older patients undergoing cardiac surgery, offering potential value for future multimodal risk stratification.

This cohort study examines the association of Alzheimer disease biomarkers with postoperative delirium in older adults undergoing elective cardiac surgery.

Abstract

Importance

Postoperative delirium (POD) is a frequent and severe complication after cardiac surgery that is associated with increased risk of dementia and adverse outcomes. Early identification of high-risk patients remains challenging.

Objective

To examine whether preoperative plasma Alzheimer disease biomarkers are associated with baseline cognition and can independently estimate POD in patients undergoing elective cardiac surgery.

Design, Setting, and Participants

This prospective, observational cohort study (FIND Delirium Risk Factors [FINDERI]) included patients 50 years or older undergoing elective cardiac surgery at a comprehensive cardiac surgery center in Germany between February 1, 2021, and October 31, 2022. Statistical analysis was performed from January to December 2025.

Exposures

Preoperative plasma concentrations of amyloid-β 1-40 (Aβ1-40), amyloid-β 42 (Aβ1-42), phosphorylated tau 181 (pTau181), and phosphorylated tau 217 (pTau217) were measured using validated immunoassays between November 1, 2022, and July 31, 2024.

Main Outcomes and Measures

The primary outcome was POD, assessed daily for 5 days postoperatively using the Confusion Assessment Method algorithm. Baseline cognitive function was assessed using the Montreal Cognitive Assessment (MoCA). Risk estimation accuracy was evaluated using receiver operating characteristic analysis and multivariate logistic regression, adjusting for relevant clinical covariates.

Results

Of 504 enrolled patients, 491 completed POD assessment (mean [SD] age, 68.4 [8.3] years; 385 [78.4%] male), of whom 106 (21.6%) developed POD. In multiple linear regression, pTau217 (β = −0.51; 95% CI, −0.89 to −0.12; P = .01) and age (β = −0.12; 95% CI, −0.16 to −0.08; P < .001) remained significant factors associated with lower MoCA scores. Patients who developed POD had higher preoperative Aβ1-40, Aβ1-42, pTau217, and Aβ1-40/Aβ1-42 × pTau217, and lower Aβ1-42/Aβ1-40. Aβ1-40 was the independent biomarker most strongly associated with POD (area under the receiver operating curve [AUC], 0.65; 95% CI, 0.59-0.71; P < .001). Combining biomarkers with clinical factors improved risk estimation (AUC, 0.74; 95% CI, 0.69-0.79; P < .001) compared with clinical factors alone (AUC, 0.73; 95% CI, 0.67-0.78; P < .001). A combined model including MoCA achieved an AUC of 0.77 (95% CI, 0.72-0.82; P < .001), driven primarily by Aβ1-40.

Conclusions and Relevance

In this prospective cohort study, preoperative Aβ1-40 showed the strongest association with POD risk among the measured biomarkers. These findings suggest that preoperative plasma Aβ1-40 may contribute to POD risk stratification, but further research is needed to validate these findings and determine clinical utility.

Introduction

Postoperative delirium (POD) is a common and serious complication after cardiac surgery, particularly in older adults, affecting up to 25% of patients.1 POD is associated with a range of adverse outcomes, including prolonged hospitalization and intensive care unit (ICU) stay, increased perioperative complications, mortality, and long-term cognitive decline.2,3,4 Established risk factors include older age, preexisting cognitive impairment, comorbidities, and various perioperative factors.5,6

Although several multivariable risk models have been proposed, most demonstrate only modest predictive accuracy and limited generalizability, and no consensus tool has yet been adopted in routine cardiac surgical practice.7,8,9,10,11,12,13,14 Concurrently, a wide array of biomarkers, particularly those reflecting inflammation, neurodegeneration, and blood brain barrier (BBB) disruption,15,16,17,18 have been investigated for their association with delirium, but no single marker has demonstrated sufficient sensitivity or specificity for individualized risk stratification.19

Blood-based biomarkers of Alzheimer disease (AD), including amyloid-β (Aβ) isoforms and phosphorylated tau (pTau), have emerged as promising candidates for capturing underlying neurodegenerative processes that may predispose patients to delirium.20,21,22 Measurable with ultrasensitive immunoassays, these biomarkers have shown strong associations with subclinical cognitive decline and AD pathology in nonsurgical populations.23,24 However, their risk estimation utility in the perioperative context—and specifically for POD—remains largely unexplored.

To address this gap, the Find Delirium Risk Factors (FINDERI) study was designed as a prospective cohort study to examine whether plasma AD biomarkers are associated with baseline cognitive function and POD risk in adults 50 years or older undergoing elective cardiac surgery.25 In this high-risk population, we evaluated whether Aβ and pTau markers improve POD risk stratification independent of traditional clinical risk factors.

Methods

Recruitment and Preoperative Assessments of the Patient Cohort

FINDERI is a prospective cohort study conducted at the University Medical Center Goettingen, Goettingen, Germany. Full details are in the published study protocol.25 Ethical approval was obtained from the Ethics Committee of the University Medical Center Goettingen, and all participants provided written informed consent. The study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)27 reporting guideline for cohort studies and the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) reporting guideline.28

Between February 1, 2021, and October 31, 2022, more than 800 patients 50 years or older undergoing elective cardiac surgery were screened. Inclusion criteria were age of 50 years or older, German language proficiency, and informed consent. Exclusion criteria were age younger than 50 years, severe cognitive impairment (eg, dementia), communication barriers (eg, aphasia or severe hearing loss), or participation in an intervention trial affecting cognitive outcomes. Baseline assessments included demographics; body mass index (BMI); cardiovascular comorbidities (eg, coronary artery disease, heart failure, atrial fibrillation, and valvular pathology); history of stroke, cancer, diabetes, kidney insufficiency, chronic obstructive pulmonary disease, thyroid disorders, depression, substance use disorder, or Parkinson disease; type and duration of cardiac surgery; and cognitive function assessed using the Montreal Cognitive Assessment (MoCA), Trail Making Test A/B, and the 6-item Cognitive Impairment Test (details previously published26).

Blood Sampling and Preanalytics

Plasma samples were collected 1 to 2 days preoperatively, processed per standard protocols,29 and stored at −80 °C. Aβ1-40, Aβ1-42, and pTau181 were measured between November 1, 2022, and July 31, 2024, using commercially available assays (Lumipulse G pTau 181 plasma, Lumipulse G β-amyloid 1-42 plasma, and Lumipulse G β-amyloid 1-40 plasma) on the fully automated Lumipulse G600II platform (Fujirebio). A new calibration curve was performed for each analyte before the experiment. Quality control analyses were performed daily using Aβ and pTau181 control samples included in the kit. Assays were performed as single measurements according to manufacturer protocols under standardized preanalytical and analytical conditions to minimize interassay variability and ensure reproducibility. pTau217 was measured using the S-Plex-pTau217 assay (Meso Scale Diagnostics), according to manufacturer instructions. All values exceeded lower limit of quantification. In AD, higher plasma ptau217 and lower Aβ1-42/Aβ1-40 ratios are blood-based biomarkers associated with AD neuropathologic change, particularly amyloid and tau pathology.30

POD Assessment

POD was assessed twice daily for 5 days postoperatively using the Confusion Assessment Method (CAM) for the ICU (CAM-ICU) and a modified version (I-CAM) for intermediate care unit and general wards, including assessment of psychomotor agitation (eg, restlessness) and retardation (eg, slowed speech).31 CAM-ICU was administered only to patients with a Richmond Agitation-Sedation Scale (RASS) score of –3 or greater. Trained investigators performed assessments using Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) criteria. Details of training and delirium algorithm are available in previous FINDERI publications.14,25

Statistical Analysis

Analyses were performed using R software, version 4.4.3 (R Foundation for Statistical Computing) from January to December 2025. We analyzed preoperative biomarkers, selected comorbidities, and intraoperative factors relevant to the biomarkers as previously described.14,26 We calculated composite biomarker terms, including the Aβ1-42/Aβ1-40 ratio, Aβ1-40/Aβ1-42 × pTau181 (AT181 term), and Aβ1-40/Aβ1-42 × pTau217 (AT217 term). Categorical variables are presented as numbers (percentages) and continuous variables as means (SDs). Group comparisons (ie, individuals with and without POD) used the Pearson χ2 test (categorical) and Welch 2-sample t test (continuous). Significance was set at 2-sided P < .05. Biomarker values were z standardized before regression. Simple logistic regression assessed associations among biomarkers, clinical variables (age, sex, MoCA, BMI, kidney failure, type 2 diabetes, coronary bypass grafting [CABG], valve surgery, other surgery [procedures other than valve surgery or CABG], and surgery duration), and POD. Kidney function was included as a binary variable due to unavailability of standardized, time-matched creatinine clearance or estimated glomerular filtration rate data. Detailed intraoperative and ICU-related variables were excluded because the aim was preoperative POD risk estimation and many of these variables occurred after the intended estimation time point.

Our primary aim was the hypothesis-driven evaluation of predefined models; thus, we did not use data-driven variable selection approaches. Multivariable logistic regression models included (1) biomarker only, (2) clinical only, (3) combined (biomarkers and clinical), (4) combined plus MoCA, (5) clinical only and Aβ1-42/Aβ1-40 ratio, (6) clinical only and AT181 term, (7) clinical only and AT217 term. Odds ratios (ORs) with 95% CIs and P values are reported (with 2-sided P < .05 indicating statistical significance).

Using simple and multiple linear regression, we also investigated the association between biomarkers and baseline cognition assessed by MoCA. Cook distance plots identified influential outliers in these linear regression models. Model fit was compared using Akaike information criteria (AIC) and bayesian information criteria (BIC).32 Multicollinearity was assessed via variance inflation factor (with >5 indicating moderate and >10 substantial multicollinearity).33 Discrimination was evaluated using receiver operating characteristic (ROC) curves (R package pROC)34 and area under the curve (AUC) (with 95% CIs and P values, R package nparcomp). Optimal cut-off points were determined by simultaneous maximization of sensitivity and specificity.35,36 Sensitivity, specificity, and positive and negative predictive values at the optimal cut-off were calculated using R package epiR. ROC curves with corresponding AUCs were generated for multiple logistic regression models. Reported AUCs were estimated in the full analysis dataset, with nearly identical values observed in a 70%-30% train-test split. Internal validation used stratified repeated 5-fold cross-validation (50 repetitions) accounting for POD class imbalance. Brier scores were used as calibration-related measures of overall risk estimation error. To exploratorily evaluate the incremental value of adding biomarkers to clinical variables and MoCA and adding MoCA to model 3 for estimating POD, we calculated continuous net reclassification improvement (NRI) values. Analyses used complete case data.

Results

Participant Flow and Baseline Characteristics

The FINDERI cohort included 491 patients (mean [SD] age, 68.4 [8.3] years; 385 [78.4%] male and 106 [21.6%] female; mean [SD] Montreal Cognitive Assessment score, 23.84 [3.62]). Of 504 patients enrolled in the FINDERI study,14 491 completed POD assessment and were included in the primary analysis. Overall, 106 patients (21.6%) had POD. Patients with POD were older and had a higher prevalence of diabetes than those without POD (Table 1). Detailed analyses of preoperative somatic and perioperative data have been reported previously.14,26

Table 1. Baseline Characteristics Stratified by POD State.

Characteristic No./total No. (%) of patientsa P valueb
Overall No POD POD
Demographics
Age, mean (SD), y 68.4 (8.3) 67.6 (8.3) 71.0 (7.7) <.001
Sex
Male 385/491 (78.4) 309/385 (80.3) 76/106 (71.7) .08
Female 106/491 (21.6) 76/385 (19.7) 30/106 (28.3)
MoCA score, mean (SD) (range, 0-30) 23.84 (3.62) 24.27 (3.37) 22.25 (4.05) <.001
BMI, mean (SD) 28.3 (4.6) 28.3 (4.6) 28.2 (4.8) .88
Comorbidities
Kidney failure 63/490 (12.9) 46/385 (11.9) 17/105 (16.2) .32
Type 2 diabetes 149/491 (30.3) 103/385 (26.8) 46/106 (43.4) .001
Type of surgery
CABG 323/491 (65.8) 255/385 (66.2) 68/106 (64.2) .78
Valve surgery 223/491 (45.4) 169/385 (43.9) 54/106 (50.9) .24
Other surgeryc 101/491 (20.6) 71/385 (18.4) 30/106 (28.3) .04
Surgery duration, mean (SD), h 4.3 (1.3) 4.2 (1.2) 4.8 (1.5) <.001
Biomarker levels, mean (SD)
Aβ1-40, pg/mL 401.0 (124.0) 389.6 (120.2) 442.5 (129.2) <.001
Aβ1-42, pg/mL 31.1 (8.7) 30.6 (8.8) 33.1 (8.5) .008
Aβ1-42/Aβ1-40 ratio 0.079 (0.010) 0.079 (0.010) 0.076 (0.010) .005
pTau181, pg/mL 1.9 (2.9) 1.8 (3.2) 2.1 (1.4) .24
AT181 term 24.5 (40.9) 23.5 (44.6) 28.1 (22.5) .15
pTau217, pg/mL 4.6 (4.7) 4.2 (3.6) 6.0 (7.2) .02
AT217 term 61.6 (71.0) 55.6 (51.4) 83.8 (115.3) .02
Open in a new tab

Abbreviations: Aβ, amyloid-β; AT181 term, Aβ1-40/Aβ1-42 × pTau181; AT217 term, Aβ1-40/Aβ1-42 × pTau217; BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); CABG, coronary artery bypass grafting; MoCA, Montreal Cognitive Assessment; POD, postoperative delirium; pTau, phosphorylated tau.

a

Unless otherwise indicated.

b

Welch 2-sample t test or Pearson χ2 test.

c

Other surgery includes surgery with procedures other than valve surgery or CABG, including combined surgery.

Biomarker Distribution and Association With POD

Biomarker data were available for 483 patients. Patients with POD had higher levels of plasma Aβ1-40 (mean [SD], 442.5 [129.2] vs 389.6 [120.2] pg/mL), Aβ1-42 (mean [SD], 33.1 [8.5] vs 30.6 [8.8] pg/mL), pTau217, and AT217 term (Figure 1) and lower Aβ1-42/Aβ1-40 ratios (mean [SD], 0.076 [0.010] vs 0.079 [0.010]).

Figure 1. Violin Plots Illustrating the Distribution of Baseline Biomarker Levels Stratified by Postoperative Delirium (POD) State and Receiver Operating Characteristic Curves Illustrating Their Discriminative Ability.

Ten-panel biomarker violin plots and R O C curves for No P O D versus P O D. Ten panels labeled A through J arranged in two rows of five. Top row panels A to E contain paired violin plots with two groups on the horizontal axis: No P O D at left in dark teal outline and P O D at right in orange outline; each violin contains a horizontal line for the median and a small dot for the mean. Panel A title A beta 1 to 40 distribution; vertical axis labeled A beta 1 to 40, p g per m L, ranging from 0 to 1250; P less than .001 printed above a bracket spanning both groups. Panel B title A beta 1 to 42 distribution; vertical axis labeled A beta 1 to 42, p g per m L, ranging from 0 to 90; P less than .01 above a spanning bracket. Panel C title A beta 42 slash A beta 40 ratio distribution; vertical axis labeled A beta 1 to 42 slash A beta 1 to 40 ratio, ranging from 0 to 0.125; P less than .01 above a spanning bracket. Panel D title p Tau 217 distribution; vertical axis labeled p Tau 217, p g per m L, ranging from 0 to 70; P less than .05 above a spanning bracket. Panel E title A T 217 term distribution; vertical axis labeled A T 217 term, log 10 scaled, with tick labels at approximately 10, 100, and 1000; P less than .05 above a spanning bracket. Bottom row panels F to J contain receiver operating characteristic plots with vertical axis Sensitivity from 0 to 1.00 and horizontal axis 1 minus Specificity from 0 to 1.00; each includes a dark teal step curve and a dotted diagonal reference line from lower left to upper right. Panel F text: A U C, 0.65, 95 percent C I, 0.59 to 0.70. Panel G text: A U C, 0.60, 95 percent C I, 0.54 to 0.66. Panel H text: A U C, 0.61, 95 percent C I, 0.55 to 0.68. Panel I text: A U C, 0.63, 95 percent C I, 0.57 to 0.69. Panel J text: A U C, 0.64, 95 percent C I, 0.58 to 0.70.

Open in a new tab

P values are derived from the Welch 2-sample t test. Aβ indicates amyloid-β; AT217 term, Aβ1-40/Aβ1-42 × pTau217; AUC, area under the receiver operating characteristic curve; pTau, phosphorylated tau. Lines within the violin plots represent the median; dots, the mean.

Association Between Biomarkers and Baseline Cognitive Function

We evaluated associations between plasma biomarkers and baseline cognitive function as assessed by MoCA. Simple linear regressions demonstrated negative associations between MoCA scores and Aβ1-40, Aβ1-42, pTau181, AT181 term, pTau217, and AT217 term (Table 2). In multiple linear regression (Table 2), pTau217 (β = −0.51; 95% CI, −0.89 to −0.12; P = .01) and age (β = −0.12; 95% CI, −0.16 to −0.08; P < .001) remained significantly associated with lower MoCA scores. However, the effect size for pTau217 appears to be associated with an influential observation in the model with pTau181. After removal, pTau217 was no longer significant (eTable 1 in Supplement 1).

Table 2. Simple and Multiple Linear Regression With Montreal Cognitive Assessment as the Dependent Variable.

Characteristic Simple Multiplea
β (95% CI) P value β (95% CI) P value VIFb
Intercept NA NA 32 (29 to 35) <.001 NA
Age −0.14 (−0.18 to −0.10) <.001 −0.12 (−0.16 to −0.08) <.001 1.1
Female sex 0.38 (−0.40 to 1.2) .34 0.59 (−0.15 to 1.3) .12 1.0
Aβ1-40 standardized biomarker −0.77 (−1.1 to −0.46) <.001 −0.27 (−0.97 to 0.44) .45 5.7
Aβ1-42 standardized biomarker −0.63 (−0.94 to −0.31) <.001 −0.10 (−0.78 to 0.57) .77 5.2
pTau181 standardized biomarker −0.42 (−0.73 to −0.11) .009 0.06 (−0.29 to 0.40) .74 1.4
pTau217 standardized biomarker −0.86 (−1.2 to −0.55) <.001 −0.51 (−0.89 to −0.12) .01 1.7
Aβ1-42/Aβ1-40 ratio standardized biomarker 0.48 (0.16 to 0.80) .003 NA NA NA
AT181 term standardized biomarker −0.45 (−0.76 to 0.13) .006 NA NA NA
AT217 term standardized biomarker −0.88 (−1.2 to −0.57) <.001 NA NA NA
Open in a new tab

Abbreviations: Aβ, amyloid β; AT181 term, Aβ1-40/Aβ1-42 × pTau181; AT217 term, Aβ1-40/Aβ1-42 × pTau217; NA, not applicable; pTau, phosphorylated tau; VIF, variance inflation factor.

a

R2 = 0.15, n = 478.

b

In the multiple logistic regression, VIFs were calculated to assess multicollinearity.

Univariate Risk Estimators of POD and ROC Analysis

Simple logistic regression identified age, MoCA, surgery duration, surgery other than CABG and valve surgery, type 2 diabetes, and several biomarkers (Aβ1-40, Aβ1-42, Aβ1-42/Aβ1-40 ratio, pTau217, AT217 term) as significantly associated with POD risk (eTable 2 in Supplement 1). ROC analyses on individual biomarkers showed that plasma Aβ1-40 had the strongest individual discriminative ability (AUC, 0.65, 95% CI, 0.59-0.71; P < .001) followed by AT181 term (AUC, 0.65; 95% CI, 0.59-0.71; P < .001), AT217 term (AUC, 0.64; 95% CI, 0.58-0.70; P < .001), pTau217 (AUC, 0.63; 95% CI, 0.57-0.69; P < .001), Aβ1-42/Aβ1-40 ratio (AUC, 0.61; 95% CI, 0.55-0.67; P < .001), and Aβ1-42 (AUC, 0.60; 95% CI, 0.54-0.66; P < .001). Optimal Aβ1-40 cutoff was 381.05 pg/mL (sensitivity, 0.62; 95% CI, 0.51-0.71; specificity, 0.62; 95% CI, 0.57-0.67; positive predictive value, 0.31; 95% CI, 0.24-0.37; and negative predictive value, 0.85; 95% CI, 0.81-0.89). Univariate ROC curves and performance metrics are given in the eFigure and eTable 3 in Supplement 1.

Multivariable Models for POD Risk Estimation

We analyzed multivariable models, including biomarkers and clinical covariates: age, sex, BMI, kidney failure, type 2 diabetes, CABG, valve surgery, other surgery, and surgery duration. In the combined model (model 3) Aβ1-40 was associated with higher POD odds (OR per SD, 1.93; 95% CI, 1.14-3.26; P = .01), with a 1-SD increase in Aβ1-40 almost doubling the odds of developing POD (Table 3). In model 5, the Aβ1-42/Aβ1-40 ratio replaced Aβ1-40 and Aβ1-42 and was inversely associated with POD (OR per SD, 0.73; 95% CI, 0.57-0.93; P = .01), suggesting lower ratios indicate higher risk (eTable 4 in Supplement 1). Model 6 (AT181 term) showed no significant association; model 7 (AT217 term) showed a positive association (OR per SD, 1.34; 95% CI, 1.04-1.78; P = .03). Full regression results are given in Table 3 (models 1-4) and eTable 4 in Supplement 1 (models 5-7); model fit and discrimination metrics are presented in eTable 5 in Supplement 1. Note that model 4 is a sensitivity analysis and is interpreted separately from models 1 to 3 and 5 to 7.

Table 3. Multiple Logistic Regression Models for Postoperative Delirium (Models 1-4).

Characteristic Biomarkers only (model 1) Clinical only (model 2) Biomarkers and clinical (model 3) Biomarkers, clinical, and MoCA (model 4)
OR (95% CI) P value VIF OR (95% CI) P value VIF OR (95% CI) P value VIF OR (95% CI) P value VIF
Intercept 0.26 (0.21-0.33) <.001 NA 0.00 (0.00-0.01) <.001 NA 0.00 (0.00-0.02) <.001 NA 0.04 (0.00-2.02) .11 NA
Female sex NA NA NA 1.87 (1.07-3.24) .03 1.1 1.98 (1.11-3.49) .02 1.1 2.20 (1.22-3.97) .008 1.1
Age NA NA NA 1.07 (1.04-1.10) <.001 1.1 1.06 (1.02-1.09) <.001 1.2 1.04 (1.01-1.08) .02 1.2
BMI NA NA NA 1.00 (0.95-1.05) .91 1.1 1.00 (0.94-1.05) .85 1.1 0.99 (0.93-1.05) .72 1.1
Kidney failure NA NA NA 1.06 (0.52-2.07) .65 1.0 0.74 (0.33-1.58) .45 1.3 0.77 (0.34-1.66) .57 1.3
Type 2 diabetes NA NA NA 2.57 (1.53-4.34) <.001 1.2 2.07 (1.20-3.57) .009 1.2 2.14 (1.23-3.74) .007 1.2
CABG NA NA NA 0.53 (0.24-1.14) .11 2.5 0.46 (0.20-1.02) .06 2.7 0.44 (0.19-1.00) .06 2.7
Valve surgery NA NA NA 0.91 (0.45-1.79) .79 2.2 0.74 (0.35-1.54) .43 2.4 0.71 (0.33-1.48) .37 2.4
Other surgerya NA NA NA 1.45 (0.81-2.56) .21 1.2 1.41 (0.77-2.55) .25 1.2 1.44 (0.78-2.63) .23 1.2
Surgery duration NA NA NA 1.63 (1.34-2.01) <.001 1.3 1.66 (1.36-2.05) <.001 1.3 1.70 (1.38-2.11) <.001 1.4
Aβ1-40 standardized biomarker 1.81 (1.12-2.93) .01 5.4 NA NA NA 1.93 (1.14-3.26) .01 5.6 1.90 (1.12-3.27) .02 5.5
Aβ1-42 standardized biomarker 0.72 (0.45-1.16) .17 5.3 NA NA NA 0.69 (0.41-1.16) .16 5.3 0.68 (0.40-1.15) .15 5.2
pTau217 standardized biomarker 1.37 (1.01-2.02) .07 2.0 NA NA NA 1.21 (0.91-1.79) .25 1.9 1.18 (0.86-1.78) .37 2.0
pTau181 standardized biomarker 0.78 (0.32-1.08) .46 2.2 NA NA NA 0.87 (0.33-1.17) .61 1.8 0.84 (0.29-1.18) .63 2.0
MoCA NA NA NA NA NA NA NA NA NA 0.89 (0.82-0.95) .001 1.1
Open in a new tab

Abbreviations: BMI, body mass index; CABG, coronary bypass grafting; MoCA, Montreal Cognitive Assessment; NA, not applicable; OR, odds ratio; pTau, phosphorylated tau; VIF, variance inflation factor.

a

Other surgery includes surgery with procedures other than valve surgery or CABG, including combined surgery.

Apparent full-sample AUCs were 0.74 (95% CI, 0.69-0.79; P < .001) for model 3 and 0.74 (95% CI, 0.68-0.79; P < .001) for model 5. In terms of model fit, model 3 achieved the lowest AIC of 448 followed by model 5 with an AIC of 449, whereas the BIC favored model 5. Brier scores ranged from 0.136 to 0.162 (eTable 5 in Supplement 1). eTable 6 in Supplement 1 reports the AUC for models 1 to 4 on a 70% training and 30% test split with preserved POD prevalence. Models 2, 3, and 4 showed comparable performance on the test set, indicating model generalizability. Repeated cross-validation showed moderate performance for the biomarker-clinical model (AUC, 0.70; Brier score, 0.15). The NRI values were 0.5151 (positive NRI, 0.2121; negative NRI, 0.3029) for adding MoCA to model 3 and 0.4083 (positive NRI, 0.0303; negative NRI, 0.3780) for adding the biomarkers to clinical variables and MoCA (eTable 7 in Supplement 1).

Across models, age, female sex, diabetes, and longer surgery duration remained associated with POD, indicating their stable contribution to POD risk. The exploratory addition of apolipoprotein ε4 carrier status did not materially improve model discrimination (eTable 8 in Supplement 1). Models combining biomarkers and clinical variables outperformed models with either alone (AUCs, 0.74; 95% CI, 0.69-0.79; P < .001 vs 0.73; 95% CI, 0.67-0.78; P < .001 for clinical only and 0.66; 95% CI, 0.60-0.72; P < .001 for biomarker only). Although the incremental improvement in model discrimination was modest (Figure 2), the consistent and independent association of Aβ1-40 with postoperative delirium across models supports its potential biological relevance.

Figure 2. Receiver Operating Characteristic Curves Comparing the Predictive Performance of Multiple Logistic Regression Models for Postoperative Delirium.

R O C curves: Sensitivity versus 1 minus Specificity for four models. Single receiver operating characteristic plot with a white background and light gray horizontal gridlines. Vertical axis label Sensitivity, ranging from zero at the bottom to one point zero zero at the top, with labeled ticks at zero point two five, zero point five zero, zero point seven five, and one point zero zero. Horizontal axis label 1 minus Specificity, ranging from zero at the left to one point zero zero at the right, with labeled ticks at zero point two five, zero point five zero, zero point seven five, and one point zero zero. Four stepwise curves begin near the origin at the lower left and progress toward the upper right corner. A dark blue gray curve corresponds to the legend entry Biomarker controlled for clinical variables. An orange curve corresponds to Clinical variables only. A light blue curve corresponds to Biomarker only. A green curve corresponds to Biomarker, clinical variables, and MoCA. The green curve lies above the other curves through much of the left and middle portions of the plot, including around 1 minus Specificity near zero point two five where Sensitivity is roughly zero point six to zero point seven, and around zero point five where Sensitivity is roughly zero point eight five to zero point nine. The light blue curve remains lower than the others in the midrange, including around 1 minus Specificity near zero point five where Sensitivity is roughly zero point seven to zero point seven five, then rises toward the upper right. The dark blue gray and orange curves track closely, with multiple crossings and small separations, and both approach Sensitivity near one point zero as 1 minus Specificity approaches one point zero. A boxed legend sits in the lower center of the plotting area with colored line samples matching each curve.

Open in a new tab

MoCA indicates Montreal Cognitive Assessment.

Model Including Baseline Cognition

We assessed whether baseline cognition influenced biomarker associations (Table 3, model 4). Aβ1-40 remained significant after adjusting for MoCA (model 3: OR per SD, 1.93; 95% CI, 1.14-3.26; model 4: OR per SD, 1.90; 95% CI, 1.12-3.27). The model demonstrated the best overall fit (AIC, 435; BIC, 497) and discrimination ability (AUC, 0.77; 95% CI: 0.72-0.82; P < .001), with variance inflation factors of 1.1 to 5.5, indicating no problematic multicollinearity (eTable 5 in Supplement 1). MoCA alone showed moderate discrimination for POD (AUC, 0.65; 95% CI, 0.59-0.71). Thus, baseline cognition contributed to risk estimation but did not explain the Aβ1-40 association.

Discussion

In this prospective cohort of older adults undergoing cardiac surgery, preoperative plasma Aβ1-40 showed the strongest biomarker association with POD, independent of baseline cognitive performance, whereas plasma pTau217 was not significant in multivariable models. These findings suggest that Aβ1-40 may reflect perioperative brain vulnerability through pathways not captured by cognitive screening or tau-related neurodegeneration.

Blood-based AD biomarkers are emerging as scalable tools for detecting early neurodegenerative changes.23,37 Prior work has linked plasma pTau217 and composite biomarker profiles to cognitive decline,37 but their associations with POD have been inconsistent.15,38 Our findings extend this literature suggesting differential associations of amyloid and tau biomarkers, which may be partly explained by the specific characteristics of our cardiac surgery population with a mean age of 68.4 years and high prevalence of vascular comorbidities but no overt dementia.39 Although pTau217 is among the most accurate biomarkers for AD neuropathologic change, the stronger association of Aβ1-40 with POD may reflect vascular and neurovascular mechanisms particularly relevant in cardiac surgery, although technical factors, including platform differences, cannot be excluded.

Plasma Aβ1-40 showed moderate discrimination for POD, consistent with prior studies20,40 in critically ill and ICU populations. No individual biomarker has shown sufficient discrimination for clinical delirium estimation, as confirmed by a recent systematic review.19 Although Aβ1-40 outperformed other markers in our cohort, its modest performance and limited positive predictive value limit its utility as a standalone screening tool.

Model 4, including biomarkers, clinical variables, and MoCA, achieved moderate discrimination (AUC, 0.77; 95% CI, 0.72-0.82), compared with the clinical-only model and MoCA alone, comparable to previously reported prospective cardiac surgery POD models.41,42 Given its moderate discrimination and limited incremental gain, the model is not suitable for routine clinical decision-making and requires external validation. NRI findings were exploratory and suggest only a limited gain in estimation improvement from adding biomarkers to MoCA and clinical variables. This pattern is consistent with studies in non–cardiac surgical populations, which report that perioperative biomarker trajectories or composite ratios may better estimate POD risk than static baseline levels.43,44,45 Similarly, spinal fluid biomarkers studies, including Aβ42, total tau, and phosphorylated tau (primarily pTau181), have produced inconsistent associations with delirium.38,46,47,48,49,50 The aim of this study was not to develop a standalone prediction tool but to evaluate whether blood-based AD biomarkers were associated with POD beyond clinical and cognitive risk factors. The modest discrimination gain after adding plasma Aβ1-40 supports a potential complementary contribution to risk stratification rather than replacement of cognitive screening. Aβ1-40 is linked to vascular Aβ deposition, cerebral amyloid angiopathy, endothelial dysfunction, and vascular comorbidities common in older surgical patients.51,52,53,54 These processes are associated with chronic, low-grade BBB dysfunction and impaired Aβ clearance. The observed association between preoperative Aβ1-40 levels and POD may reflect preexisting neurovascular vulnerability55 rather than acute BBB disruption.56,57 This interpretation remains hypothesis-generating. Together, these data support the hypothesis that Aβ1-40 may reflect neurovascular reserve, with higher levels potentially indicating impaired cerebral perfusion and metabolic resilience.

Consistent with previous studies, lower MoCA scores were associated with elevated pTau217 and greater POD risk. The persistent association between plasma Aβ1-40 and POD after adjustment for MoCA supports that Aβ1-40 may reflect vascular or BBB-related vulnerability not captured by cognitive testing. Female sex was also associated with POD, consistent with other cardiac surgery studies, and may relate to hormonal modulation of neuroinflammation, differences in cognitive reserve, sex-specific vascular pathology, or specific surgical aspects such as longer surgery duration in patients with complex cardiac anatomy.58,59

Recent research supports multimodal approaches for delirium risk stratification, with increasing evidence that integrating panels of molecular markers across inflammatory, vascular, and neurodegenerative pathways with clinical risk factors may enhance predictive accuracy over single-pathway biomarker approaches.60 Smaller cardiac surgery studies have reported improved prediction after integrating proteomic signatures and clinical data.58

Intervention trials guided by biomarker- or risk model–based patient stratification are still in the early stages. The ENGAGES trial, although innovative in its use of electroencephalography-guided anesthesia to target mechanistic risk, did not use blood biomarker–driven interventions.61 Trials should test whether biomarker-informed risk assessment improves delirium and cognitive outcomes, reflecting a broader shift toward integrated precision medicine strategies.62

Strenghths and Limitations

To our knowledge, this is the largest prospective study examining plasma Aβ1-40 and pTau217 in relation to POD in older adults undergoing cardiac surgery. Strengths include a well-characterized cohort, standardized cognitive and POD assessments, validated immunoassays, and rigorous multivariable modeling. Limitations of this study include the single-center design, lack of longitudinal biomarker data, assay or platform variability, no neurofilament light chain measurement, and limited perioperative data, which may limit generalizability. Apolipoprotein ε4 carrier status was explored but did not improve model performance. Detailed intraoperative variables, postoperative ICU factors, and neuroimaging data were not available and may have contributed to POD risk beyond surgery type and duration. Findings should be interpreted within the cardiac surgery context and require external validation before broader clinical application. Internal validation showed moderate performance (AUC, 0.70; Brier score, 0.15), and external validation is essential for clinical translation. This will be attempted through a planned multicenter trial, informed by the ongoing PREVENT-POD feasibility study.63

Conclusions

In this prospective, observational cohort study of older adults undergoing elective cardiac surgery, preoperative plasma Aβ1-40 showed the strongest biomarker association with POD and provided modest incremental predictive value beyond clinical and cognitive assessment. These findings support Aβ1-40 as a marker of neurobiological vulnerability rather than as a standalone prediction tool. Future work should validate these findings, incorporating longitudinal biomarker dynamics and advancing multimodal prediction models for risk stratification and targeted prevention strategies.

Supplement 1.

eTable 1. Multiple linear regression with MoCA as dependent variable without influential observation (N = 477)

eTable 2. Simple logistic regressions for POD

eTable 3. Area under the curve and prognostic accuracy at optimal cutoff values for POD state

eTable 4. Multiple logistic regressions for POD – Model 5-7

eTable 5. Model fit statistics for multivariable logistic regression models predicting postoperative delirium

eTable 6. AUCs of models 1-4 using a training/test split with a preserved prevalence of POD (approx. 21%) in both sets

eTable 7. Exploratory category-free net reclassification improvement (NRI)

eTable 8. Exploratory APOE ε4 analysis for postoperative delirium prediction

eFigure. ROC curves for biomarkers in predicting POD

jamanetwopen-e2625730-s001.pdf (610.5KB, pdf)
Supplement 2.

Nonauthor Collaborators. The FIND Delirium Risk Factors (FINDERI) investigators

jamanetwopen-e2625730-s002.pdf (215.7KB, pdf)
Supplement 3.

Data Sharing Statement

jamanetwopen-e2625730-s003.pdf (16.3KB, pdf)

References

  • 1.Shamsi T, Janga SR, Baskaran NU, et al. Temporal trends and severity of postoperative delirium in cardiac surgery: insights from a systematic review and meta-analysis. J Cardiothorac Vasc Anesth. 2025;39(9):2424-2435. doi: 10.1053/j.jvca.2025.05.020 [DOI] [PubMed] [Google Scholar]
  • 2.Huang H, Li H, Zhang X, et al. Association of postoperative delirium with cognitive outcomes: a meta-analysis. J Clin Anesth. 2021;75:110496. doi: 10.1016/j.jclinane.2021.110496 [DOI] [PubMed] [Google Scholar]
  • 3.Kunicki ZJ, Ngo LH, Marcantonio ER, et al. Six-year cognitive trajectory in older adults following major surgery and delirium. JAMA Intern Med. 2023;183(5):442-450. doi: 10.1001/jamainternmed.2023.0144 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Au E, Thangathurai G, Saripella A, et al. Postoperative outcomes in elderly patients undergoing cardiac surgery with preoperative cognitive impairment: a systematic review and meta-analysis. Anesth Analg. 2023;136(6):1016-1028. doi: 10.1213/ANE.0000000000006346 [DOI] [PubMed] [Google Scholar]
  • 5.Ormseth CH, LaHue SC, Oldham MA, Josephson SA, Whitaker E, Douglas VC. Predisposing and precipitating factors associated with delirium: a systematic review. JAMA Netw Open. 2023;6(1):e2249950. doi: 10.1001/jamanetworkopen.2022.49950 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Koster S, Hensens AG, Schuurmans MJ, van der Palen J. Risk factors of delirium after cardiac surgery: a systematic review. Eur J Cardiovasc Nurs. 2011;10(4):197-204. doi: 10.1016/j.ejcnurse.2010.09.001 [DOI] [PubMed] [Google Scholar]
  • 7.Gunther M, Pipolo D, Jiang S, et al. Stanford study of the validity of the pre-interventional preventive risk assessment (PIPRA) tool for at-risk elderly patients undergoing elective orthopedic surgery. J Psychosom Res. 2025;195:112196. doi: 10.1016/j.jpsychores.2025.112196 [DOI] [PubMed] [Google Scholar]
  • 8.Jauk S, Kramer D, Sumerauer S, Veeranki SPK, Schrempf M, Puchwein P. Machine learning-based delirium prediction in surgical in-patients: a prospective validation study. JAMIA Open. 2024;7(3):ooae091. doi: 10.1093/jamiaopen/ooae091 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Shaw KM, Shao YP, Ghanta M, et al. Daily automated prediction of delirium risk in hospitalized patients: model development and validation. JMIR Med Inform. 2025;13:e60442. doi: 10.2196/60442 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Eschweiler GW, Czornik M, Herrmann ML, et al. Presurgical screening improves risk prediction for delirium in elective surgery of older patients: the PAWEL RISK Study. Front Aging Neurosci. 2021;13:679933. doi: 10.3389/fnagi.2021.679933 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Lindroth H, Bratzke L, Twadell S, et al. Predicting postoperative delirium severity in older adults: the role of surgical risk and executive function. Int J Geriatr Psychiatry. 2019;34(7):1018-1028. doi: 10.1002/gps.5104 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Menzenbach J, Kirfel A, Guttenthaler V, et al. ; PROPDESC Collaboration Group . PRe-Operative Prediction of postoperative DElirium by appropriate SCreening (PROPDESC) development and validation of a pragmatic POD risk screening score based on routine preoperative data. J Clin Anesth. 2022;78:110684. doi: 10.1016/j.jclinane.2022.110684 [DOI] [PubMed] [Google Scholar]
  • 13.Rudolph JL, Jones RN, Levkoff SE, et al. Derivation and validation of a preoperative prediction rule for delirium after cardiac surgery. Circulation. 2009;119(2):229-236. doi: 10.1161/CIRCULATIONAHA.108.795260 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Sadlonova M, Hansen N, Esselmann H, et al. Preoperative delirium risk screening in patients undergoing a cardiac surgery: results from the prospective observational FINDERI study. Am J Geriatr Psychiatry. 2024;32(7):835-851. doi: 10.1016/j.jagp.2023.12.017 [DOI] [PubMed] [Google Scholar]
  • 15.Ballweg T, White M, Parker M, et al. Association between plasma tau and postoperative delirium incidence and severity: a prospective observational study. Br J Anaesth. 2021;126(2):458-466. doi: 10.1016/j.bja.2020.08.061 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Fong TG, Vasunilashorn SM, Kivisäkk P, et al. Biomarkers of neurodegeneration and neural injury as potential predictors for delirium. Int J Geriatr Psychiatry. 2024;39(1):e6044. doi: 10.1002/gps.6044 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Breitkopf M, Ahrens E, Herrmann ML, et al. Preoperative hypoxic biomarkers and postoperative delirium in patients with obstructive sleep apnea. J Anesth. 2025;39(1):31-40. doi: 10.1007/s00540-024-03417-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Noah AM, Almghairbi D, Evley R, Moppett IK. Preoperative inflammatory mediators and postoperative delirium: systematic review and meta-analysis. Br J Anaesth. 2021;127(3):424-434. doi: 10.1016/j.bja.2021.04.033 [DOI] [PubMed] [Google Scholar]
  • 19.Aldecoa C, Bettelli G, Bilotta F, et al. Update of the European Society of Anaesthesiology and Intensive Care Medicine evidence-based and consensus-based guideline on postoperative delirium in adult patients. Eur J Anaesthesiol. 2024;41(2):81-108. doi: 10.1097/EJA.0000000000001876 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Chan CK, Song Y, Greene R, et al. Meta-analysis of ICU delirium biomarkers and their alignment with the NIA-AA Research Framework. Am J Crit Care. 2021;30(4):312-319. doi: 10.4037/ajcc2021771 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Palmqvist S, Whitson HE, Allen LA, et al. Alzheimer’s Association Clinical Practice Guideline on the use of blood-based biomarkers in the diagnostic workup of suspected Alzheimer’s disease within specialized care settings. Alzheimers Dement. 2025;21(7):e70535. doi: 10.1002/alz.70535 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Liang F, Baldyga K, Quan Q, et al. Preoperative plasma Tau-PT217 and Tau-PT181 are associated with postoperative delirium. Ann Surg. 2023;277(6):e1232-e1238. doi: 10.1097/SLA.0000000000005487 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Palmqvist S, Stomrud E, Cullen N, et al. An accurate fully automated panel of plasma biomarkers for Alzheimer’s disease. Alzheimers Dement. 2023;19(4):1204-1215. doi: 10.1002/alz.12751 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Morgado B, Klafki HW, Bauer C, et al. Assessment of immunoprecipitation with subsequent immunoassays for the blood-based diagnosis of Alzheimer’s disease. Eur Arch Psychiatry Clin Neurosci. 2025;275(8):2215-2227. doi: 10.1007/s00406-023-01751-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Sadlonova M, Vogelgsang J, Lange C, et al. Identification of risk factors for delirium, cognitive decline, and dementia after cardiac surgery (FINDERI-find delirium risk factors): a study protocol of a prospective observational study. BMC Cardiovasc Disord. 2022;22(1):299. doi: 10.1186/s12872-022-02732-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Itting PT, Sadlonova M, Santander MJ, et al. ; FINDERI investigators . Intra- and early postoperative predictors of delirium risk in cardiac surgery: results from the prospective observational FINDERI study. Int J Surg. 2025;111(4):2872-2885. doi: 10.1097/JS9.0000000000002265 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative . The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: guidelines for reporting observational studies. Int J Surg. 2014;12(12):1495-1499. doi: 10.1016/j.ijsu.2014.07.013 [DOI] [PubMed] [Google Scholar]
  • 28.Collins GS, Moons KGM, Dhiman P, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. doi: 10.1136/bmj-2023-078378 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Vogelgsang J, Hansen N, Stark M, et al. Plasma amyloid beta X-42/X-40 ratio and cognitive decline in suspected early and preclinical Alzheimer’s disease. Alzheimers Dement. 2024;20(8):5132-5142. doi: 10.1002/alz.13909 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Jack CR Jr, Andrews JS, Beach TG, et al. Revised criteria for diagnosis and staging of Alzheimer’s disease: Alzheimer’s Association Workgroup. Alzheimers Dement. 2024;20(8):5143-5169. doi: 10.1002/alz.13859 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Thomas C, Kreisel SH, Oster P, Driessen M, Arolt V, Inouye SK. Diagnosing delirium in older hospitalized adults with dementia: adapting the confusion assessment method to international classification of diseases, tenth revision, diagnostic criteria. J Am Geriatr Soc. 2012;60(8):1471-1477. doi: 10.1111/j.1532-5415.2012.04066.x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Dziak JJ, Coffman DL, Lanza ST, Li R, Jermiin LS. Sensitivity and specificity of information criteria. Brief Bioinform. 2020;21(2):553-565. doi: 10.1093/bib/bbz016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Kim JH. Multicollinearity and misleading statistical results. Korean J Anesthesiol. 2019;72(6):558-569. doi: 10.4097/kja.19087 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Robin X, Turck N, Hainard A, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics. 2011;12:77. doi: 10.1186/1471-2105-12-77 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Konietschke F, Placzek M, Schaarschmidt F, Hothorn LA. nparcomp: an R software package for nonparametric multiple comparisons and simultaneous confidence intervals. J Stat Softw. 2015;64(9). doi: 10.18637/jss.v064.i09 [DOI] [Google Scholar]
  • 36.López-Ratón M, Rodríguez-Álvarez MX, Suárez CC, Sampedro FG. OptimalCutpoints: an R package for selecting optimal cutpoints in diagnostic tests. J Stat Softw. 2014;61(8). doi: 10.18637/jss.v061.i08 [DOI] [Google Scholar]
  • 37.Palmqvist S, Tideman P, Mattsson-Carlgren N, et al. Blood biomarkers to detect Alzheimer disease in primary care and secondary care. JAMA. 2024;332(15):1245-1257. doi: 10.1001/jama.2024.13855 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Fong TG, Vasunilashorn SM, Gou Y, et al. ; RISE Study Group . Association of CSF Alzheimer’s disease biomarkers with postoperative delirium in older adults. Alzheimers Dement (N Y). 2021;7(1):e12125. doi: 10.1002/trc2.12125 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.McKay TB, Smith M, Mueller A, et al. Associations between Aβ40, Aβ42, and tau and postoperative delirium in older adults undergoing cardiac surgery. J Neurol. 2025;272(6):393. doi: 10.1007/s00415-025-13145-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Simons KS, van den Boogaard M, Hendriksen E, et al. Temporal biomarker profiles and their association with ICU acquired delirium: a cohort study. Crit Care. 2018;22(1):137. doi: 10.1186/s13054-018-2054-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Chen H, Yu D, Zhang J, Li J. Machine learning for prediction of postoperative delirium in adult patients: a systematic review and meta-analysis. Clin Ther. 2024;46(12):1069-1081. doi: 10.1016/j.clinthera.2024.09.013 [DOI] [PubMed] [Google Scholar]
  • 42.Tu Y, Zhu H, Zhang X, Huang S, Tu W. Machine Learning-Based prediction models for postoperative delirium: a systematic review and meta-analysis. BMC Psychiatry. 2025;25(1):940. doi: 10.1186/s12888-025-07401-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Fan Q, Wang Y, Lu Z, et al. Association between plasma amyloid-beta 42 ratio and postoperative delirium in elderly patients undergoing major abdominal surgery: secondary analysis of a randomized controlled trial. Brain Behav. 2025;15(4):e70501. doi: 10.1002/brb3.70501 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Payne T, Taylor J, Kunkel D, et al. Association of preoperative to postoperative change in cerebrospinal fluid fibrinogen with postoperative delirium. BJA Open. 2024;12:100349. doi: 10.1016/j.bjao.2024.100349 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Parker M, White M, Casey C, et al. Cohort analysis of the association of delirium severity with cerebrospinal fluid amyloid-tau-neurodegeneration pathologies. J Gerontol A Biol Sci Med Sci. 2022;77(3):494-501. doi: 10.1093/gerona/glab203 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Xie Z, Swain CA, Ward SAP, et al. Preoperative cerebrospinal fluid β-amyloid/tau ratio and postoperative delirium. Ann Clin Transl Neurol. 2014;1(5):319-328. doi: 10.1002/acn3.58 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Geng J, Zhang Y, Chen H, et al. Associations between Alzheimer’s disease biomarkers and postoperative delirium or cognitive dysfunction: a meta-analysis and trial sequential analysis of prospective clinical trials. Eur J Anaesthesiol. 2024;41(3):234-244. doi: 10.1097/EJA.0000000000001933 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Cunningham EL, McGuinness B, McAuley DF, et al. CSF beta-amyloid 1-42 concentration predicts delirium following elective arthroplasty surgery in an observational cohort study. Ann Surg. 2019;269(6):1200-1205. doi: 10.1097/SLA.0000000000002684 [DOI] [PubMed] [Google Scholar]
  • 49.Idland AV, Wyller TB, Støen R, et al. Preclinical amyloid-β and axonal degeneration pathology in delirium. J Alzheimers Dis. 2017;55(1):371-379. doi: 10.3233/JAD-160461 [DOI] [PubMed] [Google Scholar]
  • 50.Hall RJ, Watne LO, Cunningham E, et al. CSF biomarkers in delirium: a systematic review. Int J Geriatr Psychiatry. 2018;33(11):1479-1500. doi: 10.1002/gps.4720 [DOI] [PubMed] [Google Scholar]
  • 51.Toledo JB, Vanderstichele H, Figurski M, et al. ; Alzheimer’s Disease Neuroimaging Initiative . Factors affecting Aβ plasma levels and their utility as biomarkers in ADNI. Acta Neuropathol. 2011;122(4):401-413. doi: 10.1007/s00401-011-0861-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Yamazaki Y, Shinohara M, Shinohara M, et al. Selective loss of cortical endothelial tight junction proteins during Alzheimer’s disease progression. Brain. 2019;142(4):1077-1092. doi: 10.1093/brain/awz011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Alonzo NC, Hyman BT, Rebeck GW, Greenberg SM. Progression of cerebral amyloid angiopathy: accumulation of amyloid-beta40 in affected vessels. J Neuropathol Exp Neurol. 1998;57(4):353-359. doi: 10.1097/00005072-199804000-00008 [DOI] [PubMed] [Google Scholar]
  • 54.Syrjanen JA, Campbell MR, Algeciras-Schimnich A, et al. Associations of amyloid and neurodegeneration plasma biomarkers with comorbidities. Alzheimers Dement. 2022;18(6):1128-1140. doi: 10.1002/alz.12466 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Sapkota S, Erickson K, Fletcher E, Tomaszewski Farias SE, Jin LW, DeCarli C; Alzheimer’s Disease Neuroimaging Initiative* . Vascular risk predicts plasma amyloid β 42/40 through cerebral amyloid burden in apolipoprotein e ε4 carriers. Stroke. 2023;54(5):1227-1235. doi: 10.1161/STROKEAHA.122.041854 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Sun L, Jia P, Zhang J, et al. Production of inflammatory cytokines, cortisol, and Aβ1-40 in elderly oral cancer patients with postoperative delirium. Neuropsychiatr Dis Treat. 2016;12:2789-2795. doi: 10.2147/NDT.S113077 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Taylor J, Parker M, Casey CP, et al. Postoperative delirium and changes in the blood-brain barrier, neuroinflammation, and cerebrospinal fluid lactate: a prospective cohort study. Br J Anaesth. 2022;129(2):219-230. doi: 10.1016/j.bja.2022.01.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Gonçalves MCB, Khera T, Otu HH, et al. Multivariable predictive model of postoperative delirium in cardiac surgery patients: proteomic and demographic contributions. Anesth Analg. 2025;140(2):476-487. doi: 10.1213/ANE.0000000000007293 [DOI] [PubMed] [Google Scholar]
  • 59.Wiredu K, Mueller A, McKay TB, Behera A, Shaefi S, Akeju O. Sex differences in the incidence of postoperative delirium after cardiac surgery: a pooled analyses of clinical trials. Anesthesiology. 2023;139(4):540-542. doi: 10.1097/ALN.0000000000004656 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Wiredu K, Aduse-Poku E, Shaefi S, Gerber SA. Proteomics for the discovery of clinical delirium biomarkers: a systematic review of major studies. Anesth Analg. 2023;136(3):422-432. doi: 10.1213/ANE.0000000000006246 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Wildes TS, Mickle AM, Ben Abdallah A, et al. ; ENGAGES Research Group . Effect of electroencephalography-guided anesthetic administration on postoperative delirium among older adults undergoing major surgery: the ENGAGES randomized clinical trial. JAMA. 2019;321(5):473-483. doi: 10.1001/jama.2018.22005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Fong TG, Inouye SK. The inter-relationship between delirium and dementia: the importance of delirium prevention. Nat Rev Neurol. 2022;18(10):579-596. doi: 10.1038/s41582-022-00698-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.German Clinical Trials Register. Feasibility, acceptance and preliminary efficacy of the multimodal AKTIVER heart delirium prevention program in patients following cardiac surgery: randomized controlled PREVENT POstoperative Delirium (PREVENT-POD) study. DRKS00035091. Accessed June 11, 2025. https://drks.de/search/en/trial/DRKS00035091

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement 1.

eTable 1. Multiple linear regression with MoCA as dependent variable without influential observation (N = 477)

eTable 2. Simple logistic regressions for POD

eTable 3. Area under the curve and prognostic accuracy at optimal cutoff values for POD state

eTable 4. Multiple logistic regressions for POD – Model 5-7

eTable 5. Model fit statistics for multivariable logistic regression models predicting postoperative delirium

eTable 6. AUCs of models 1-4 using a training/test split with a preserved prevalence of POD (approx. 21%) in both sets

eTable 7. Exploratory category-free net reclassification improvement (NRI)

eTable 8. Exploratory APOE ε4 analysis for postoperative delirium prediction

eFigure. ROC curves for biomarkers in predicting POD

jamanetwopen-e2625730-s001.pdf (610.5KB, pdf)
Supplement 2.

Nonauthor Collaborators. The FIND Delirium Risk Factors (FINDERI) investigators

jamanetwopen-e2625730-s002.pdf (215.7KB, pdf)
Supplement 3.

Data Sharing Statement

jamanetwopen-e2625730-s003.pdf (16.3KB, pdf)

Articles from JAMA Network Open are provided here courtesy of American Medical Association

RESOURCES