# paper.pdf ## Metadata - PDFFormatVersion=1.6 - Language=en - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260403105502+05'30' - Creator=LaTeX with hyperref package - ModDate=D:20260403105535+05'30' - Producer=Acrobat Distiller 26.0 (Windows) - Title=Neutrophil inflammation metrics are associated with the risk of future dementia in large data from NYU Langone Hospitals and the Veterans Health Administration - Custom.WPS-ARTICLEDOI=10.1002/alz.71335 - Custom.WPS-JOURNALDOI=10.1002/(ISSN)1552-5279 - Custom.WPS-PROCLEVEL=3 - xmp:createdate=2015-08-18T08:05:36+05:30 - xmp:modifydate=2016-06-21T14:17:29-04:00 - xmp:metadatadate=2016-06-21T14:17:29-04:00 - pdf:producer=Acrobat Distiller 9.0 (Windows) - pdfx:wps-articledoi=10.1002/alz.71335 - pdfx:wps-journaldoi=10.1002/(ISSN)1552-5279 - pdfx:wps-proclevel=3 - jav:journal_article_version=VoR - prism:doi=10.1002/alz.71335 - dc:format=application/pdf - dc:title=Neutrophil inflammation metrics are associated with the risk of future dementia in large data from NYU Langone Hospitals and the Veterans Health Administration - xmpmm:documentid=uuid:234b9391-712f-3d10-8427-f490be740d64 - xmpmm:instanceid=uuid:27e5ab08-9f56-392b-8ce7-263994c2fc32 ## Contents ### Page 1 Received: 28 July 2025 Revised: 10 February 2026 Accepted: 4 March 2026 DOI: 10.1002/alz.71335R E S E A R C H A R T I C L E Neutrophil inflammation metrics are associated with the risk of future dementia in large data from NYU Langone Hospitals and the Veterans Health Administration Tianshe He1,2 Rebecca A. Betensky3 Ricardo S. Osorio2,4 Kaitlin Swinnerton1 Chunlei Zheng1,5 Tovia Jacobs2 Alok Vedvyas6 Karyn Marsh6 Joshua Chodosh7 Ula Y. Hwang8 Natalia Sifnugel8 Omonigho M. Bubu2,6 Thomas Wisniewski2,6,9 Mary Brophy1 Nhan V. Do1 Nathanael R. Fillmore1,10 Jaime Ramos-Cejudo1,2,61 VA Boston Cooperative Studies Program, MAVERIC, VA Boston Healthcare System, Boston, Massachusetts, USA 2 Department of Psychiatry, New York University Grossman School of Medicine, New York, New York, USA 3 Department of Biostatistics, NYU School of Global Public Health, New York, New York, USA 4 Nathan Kline Institute, Orangeburg, New York, USA 5 Department of Medicine, Boston University School of Medicine, Boston, Massachusetts, USA 6 Center for Cognitive Neurology, Department of Neurology, New York University Grossman School of Medicine, New York, New York, USA 7 Division of Geriatrics and Palliative Care, Department of Medicine, New York University Grossman School of Medicine, New York, New York, USA 8 Department of Population Health, New York University Grossman School of Medicine, New York, New York, USA 9 Department of Pathology, New York University Grossman School of Medicine, New York, New York, USA 10 Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USACorrespondenceRamos-Cejudo, NYU Grossman School of Medicine, Division of Brain Aging, 145 East 32nd Street, New York, NY, USA. Email: Jaime.ramoscejudo@nyulangone.org Nathanael R. Fillmore. VA Boston Cooperative Studies Program, MAVERIC, VA Boston Healthcare System, 2 Lafayette Avenue, Boston, MA, USA. Email: Nathanael.fillmore@va.govFunding informationthe National Institutes of Health, Grant/Award Numbers: R01AG070821, R01AG079282, P30AG066512; National Alzheimer’s Coordinating Center AY2022/23 New Investigator Award to J.R.C., and the Cooperative Studies Program, VA Boston Healthcare System. O.M.B. is supported by theAbstract INTRODUCTION: Neutrophil-to-lymphocyte ratio (NLR), a marker of systemic inflam- mation, has been linked to dementia risk, but prior studies were limited by small sample sizes.METHODS: We assessed the association between baseline NLR and incident Alzheimer’s disease (AD) and Alzeimer’s disease and related dementias (AD/ADRD) using electronic health records from New York University (NYU) (n = 284,530) and the Veterans Health Administration [VA] (n = 85,836) Hospitals from 2011 to 2023. AD/ADRD diagnoses were identified via International Classification of Diseases (ICD) codes ≥6 months post-baseline. Cox models and cumulative incidence functions (CIFs) adjusted for demographic and clinical variables, with death as a competing risk.RESULTS: Higher NLR was associated significantly with increased AD/ADRD risk in both cohorts (NYU hazard ratio [HR] = 1.07, 95% confidence interval [CI] 1.02–1.15; Nathanael R. Fillmore and Jaime Ramos-Cejudo contributed equally as senior authors. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. © 2026 The Author(s). Alzheimer’s & Dementia published by Wiley Periodicals LLC on behalf of Alzheimer’s Association.Alzheimer’s Dement. 2026;22:e71335. wileyonlinelibrary.com/journal/alz 1 of 11https://doi.org/10.1002/alz.71335 ### Page 2 2 of 11 HE ET AL .NIH through the following grants, Grant/Award Numbers: K23AG068534, R01AG082278, RF1AG083975; Alzheimer’s Association grant, Grant/Award Number: AARG-21-848397; and BrightFocus Foundation, Grant/Award Number: A2022033S; U.S. Department of Veterans Affairs, Grant/Award Number: Cooperative Studies Program; Foundation for the National Institutes of Health, Grant/Award Numbers: K23AG068534, R01AG082278, RF1AG083975 VA HR = 1.21, 95% CI 1.10–1.34). Spline analysis further confirmed a continuous dose– response relationship, and subgroup analyses showed higher risk among female and Hispanic patients.DISCUSSION: Elevated NLR is independently associated with higher AD/ADRD risk across diverse populations, highlighting the role of systemic inflammation and neutrophil-mediated pathways in neurodegeneration.K E Y W O R D SAD, ADRD, Alzheimer’s disease, CBC, dementia, neutrophils, neutrophil-to-lymphocyte ratio, NLRHighlights∙ We studied whether the neutrophil-to-lymphocyte ratio (NLR) was associated with future risk of Alzheimer’s disease and Alzeimer’s disease (AD) and related dementias (AD/ADRD).∙ We explored two large electronic health record systems with varying demographic and clinical characteristics.∙ We used inverse probability weighting Cox models and cumulative incident function estimation in adjusted models and a large observation window.∙ Higher NLR was independently and positively associated with risk of future AD/ADRD.∙ Subgroup analyses showed that female and Hispanic participants consistently exhibited higher hazard ratios for NLR in relation to the AD/ADRD risk.1 BACKGROUNDThere is growing interest in inflammation as a potential contribu- tor of Alzheimer’s disease and Alzheimer’s disease (AD) and related dementias (AD/ADRD) pathology. In recent years, the National Insti- tute on Aging and Alzheimer’s Association (NIA-AA) incorporated the inflammation/vascular/α-synuclein (I/V/S) framework to supplement the predominant amyloid/tau/neurodegeneration (A/T/N) paradigm for AD definition. 1,2 Neutrophils are the most abundant leukocyte type in the human blood and one of the fundamental cell types involved in innate immunity, and recent studies have identified a potential con- tribution to AD pathology. 3–5 Although the exact mechanisms are not yet known, prior research reported the presence of neutrophil-related markers in human AD lesions as well as an association of neutrophil lev- els with AD molecular markers. 3,6,7 One of the main limitations when studying neutrophil functional phenotypes is the need to use fresh blood samples. Although efforts have been made to identify poten- tial proxies of neutrophil inflammation available in larger cohorts, this issue remains unresolved. The neutrophil-to-lymphocyte ratio (NLR) is an easily accessible inflammatory biomarker that has been associated with dementia risk, 8,9 cognitive decline, 10 and systemic inflammation in conditions associated with aging. 11–14 Given its broad applicability across clinical and research settings—including its potential to identify at-risk indi- viduals, track disease progression, 15 and ease of access through blood sampling, the NLR is often used as an indicator of the balance between systemic inflammation and the adaptive immune response. Although their best characterized function is pathogen defense, neutrophils are also implicated in tissue repair mechanisms. 16,17 When tissue is dam- aged and subsequently inflamed, neutrophils accumulate at the site of injury-releasing signals to attract additional immune cells and under- going apoptosis, 18 ultimately creating neutrophil extracellular traps (NETS) and resolving inflammation as part of the clearance process. 19 During normal inflammation, neutrophils are cleared from circulation after insult resolution, but neutrophil signaling may become altered during aging, compromising clearance and increasing the risk to cre- ate chronic inflammatory sites, 11–13 which are part of what has been described as inflammaging. 20 In recent years, neutrophils have been linked with AD pathology, 21 primarily through the pathological breakdown of the blood–brain barrier 6,22 in association with an increase in tau hyperphosphoryla- tion in animal models. 6 In human studies, higher NLR levels have been reported in individuals with dementia compared with controls, and a positive association between higher NLR levels and the risk of subse- quent dementia risk has been noted. 8,15,23 An increase in NLR values is also observed in other clinical conditions, raising the question of ### Page 3 HE ET AL . 3 of 11RESEARCH IN CONTEXT1. Systematic review: We reviewed the literature avail- able on PubMed that investigated the relationship between systemic inflammation, neutrophil activity, and Alzheimer’s disease (AD) and AD-related dementias (AD/ADRD). Prior studies have linked the neutrophil- to-lymphocyte ratio (NLR) to dementia and AD risk, but evidence has largely come from small or moderately sized cohorts with limited follow-up, restricted population heterogeneity, and limited control for confounding and competing risks. As a result, the generalizability and long- term relevance of NLR as a marker of future AD/ADRD risk have remained uncertain. 2. Interpretation: Using two large and independent elec- tronic health record systems, this study demonstrates that elevated NLR is consistently and independently asso- ciated with incident AD/ADRD across health care set- tings and demographic subgroups. The observed dose– response relationship and subgroup heterogeneity sup- port a scalable link between systemic inflammation and dementia risk, reinforcing inflammatory and immune pathways as relevant contributors to neurodegeneration. 3. Future directions: Future studies should integrate lon- gitudinal NLR trajectories with molecular biomarkers, as well as clarify causal mechanisms linking neutrophil biology to neurodegeneration, and determining whether targeting systemic inflammation can meaningfully modify dementia risk or progression. whether the NLR is directly linked with AD pathology or the result of underlying comorbidities. 4,8,24–27 The inconsistent results on this topic in previous studies are due partly to the difficulty in separating the effects of age and chronic conditions like diabetes and hypertension, which are known to raise NLR levels. 28–40 To this end, propensity score weighing and competing-risks analysis have been used previously to examine the association of the NLR with future dementia, 9 but studies have been limited by small sample sizes, thereby compromising sta- tistical power. Whether such an association can be generalized, and pervades in the long-term, still needs to be determined. The aim of this work is to investigate whether the association of NLR levels and risk of AD/ADRD is generalizable and scalable across large and comprehensive health care systems. We studied patients at New York University (NYU) Langone Hospitals (NYULH) and the Veterans Health Administration (VHA or VA) Hospitals who are potentially at risk of future AD/ADRD. Capitalizing on the relatively large and diverse datasets, we conducted extensive subgroup analyses to further establish the associations and determined whether certain demographic subsets are particularly susceptible to AD/ADRD rates through NLR levels.2 METHODS 2.1 Study design and inclusion criteriaThe present study follows the reporting guidelines of the Strengthen- ing the Reporting of Observational Studies in Epidemiology (STROBE) group. Our study was approved under NYU Institutional Review Board (IRB) s23-00860, “incidence, risk and progression of neurodegenera- tion in NYULH data,” and by the VA Boston Research and Development committee as an exempt study prior to data collection and analysis with a waiver of informed consent due to the use of existing data, per the Common Rule. The datasets comprise medical records from NYULH and the VHA and are not publicly accessible. Researchers may request access to a limited, de-identified dataset, contingent upon the ratio- nality of the request and approval from local and national policies and ethics boards. NYU Langone data were obtained by the NYU Data Core team and contain comprehensive clinical information, demographics, Interna- tional Classification of Diseases, Tenth Revision (ICD-10) diagnoses, procedures, and laboratory data across all six of the NYU hospitals. For the VA portion of our study, we used patient-level data from the US Department of Veterans Affairs Corporate Data Warehouse (CDW), a centralized and nationwide electronic health records (EHR) database of VA patients containing comprehensive clinical information, demo- graphics, Ninth and Tenth Revision ICD diagnosis (ICD-9 and ICD-10), procedures, prescription fills, and laboratory data. 41 ICD-9 (before October 2015) and ICD-10 (after October 2015) codes were used to obtain diagnosis information. For both NYU and VA data, the observa- tion window for all covariates, exposures, and outcomes in our primary analysis was from January 1, 2011 to December 31, 2023. The full list of ICD codes used for each data source is included in Table S1. Our inclusion criteria are: (1) the presence of neutrophil and lym- phocyte total counts and the associated timestamp that corresponds to the baseline time origin; (2) age >55 at baseline, which was applied in order to enrich our sample with those at potential higher risk of devel- oping AD/ADRD within the study window; and (3) not having received a diagnosis of AD/ADRD at any time before origin. To reduce the possi- bility of capturing imminent outcomes not controlled under our study design, only outcomes occurring at least 6 months after the clinical lab acquisition (baseline) were considered. We stipulated an additional requirement for the VA cohort that patients must have continuous follow-up visits at least once a year. However, this was not possible within the NYU hospitals, as it is an open system where patients can switch in and out of continuous care.2.2 Outcome definition: AD/ADRDOur primary outcome was the time to incident diagnosis of AD/ADRD, defined using the Million Veteran Program (MVP) Cognitive Decline and Dementia During Aging Working Group criteria. Patients are iden- tified as having AD/ADRD if they receive two positive ICD codes for ### Page 4 4 of 11 HE ET AL .either AD, non-specific dementia, or related dementias such as vas- cular dementia or Lewy body dementia. The outcome definition in our study is derived from those used in previously published stud- ies using ICD-based dementia diagnosis. 42,43 Time-to-event responses were used for survival analysis (see Statistical Analysis).2.3 Primary exposure: Neutrophil-to-lymphocyte ratio (NLR)The primary variable of interest in our study was the first available NLR for a patient satisfying the inclusion criteria (e.g., after the age of 55 and above, and January 1, 2011), estimated by dividing the total (absolute) counts of neutrophils/mL to lymphocytes/mL. NLR values were scaled by standard deviation (SD) across each cohort. Untransformed NLR and log-transformed NLR were also evaluated for sensitivity purposes. In both cases, the NLR values were obtained as part of the standard clinical labs (e.g. complete blood cell count panels [complete blood count (CBC)]). Neutrophil and lymphocyte measurements were demarcated as such through Logical Observa- tion Identifier Names and Codes (LOINC) codes 751-8 for neutrophils and 731-0 for lymphocytes per volume (by automated count) in both NYU and VA cohorts. Outliers for NLRs were filtered by removing the top 1% quantile of NLR, since preliminary exploration indicated that this threshold removed potential cases of neutrophilia and mis- labeled units or values. For patients whose first NLR was flagged as an outlier, but who had subsequent non-outlier measurements, the first non-outlier NLR was used as the baseline value for the patient.2.4 CovariatesCovariates included demographic and medical history (MH) variables. Demographic variables included: age at the time of the lab result (cal- culated using date of birth), self-reported race (White, Black or African American, Asian, or Other), ethnicity (Hispanic or non-Hispanic), and sex (male or female). MH covariates were conditions known to be risk factors for AD/ADRD derived from all available EHR data before baseline within the study window (2011–2023) and may temporally precede the primary exposure. MH covariates were modeled as a binary ever/never indicators before exposure and included history of diabetes, hypertension, cardio- and cerebrovascular disease (CVD), traumatic brain injury (TBI), and hypercholesterolemia, all of which defined using ICD-9 and ICD-10 (Table S1). CVD was defined as a composite indicator and comprised peripheral artery disease (PVD), coronary artery disease (CAD), myocardial infarct, heart failure, and stroke. We did not exclude patients with inflammatory conditions (e.g., rheumatoid arthritis) as such cases may be informative toward the hypothesized NLR-AD/ADRD relationship and the subject of future research.2.5 Statistical analysisWe tested the association of NLR with AD/ADRD in two separate ways. First, we assessed whether the elevated NLR, when dichotomized as high or low (1 if > median, 0 if ≤ median), was independently asso- ciated with AD/ADRD risk. We used cause-specific Cox proportional hazard (PH) regression with elevated NLR as the primary exposure of interest. We then evaluated whether elevated NLR levels were associ- ated with the probability of AD/ADRD using the cumulative incidence function (CIF). Second, we examined the dose–response relationship by modeling continuous NLR values using cubic spline terms in a Cox model. As an exploratory analysis to assess the contributions of NLR as nonlinear spline terms, we applied random forest survival methods (using R package ranger) to estimate variable importance after adjust- ing for covariates. In the Cox and random forest analyses, deaths were censored. The CIF analysis treated death as a competing event. For all analyses, two-sided p-values below 0.05 were considered statistically significant. Data analyses were performed in separate environments, with NYU analyses done within the virtual statistical network interface at NYU Langone and VA analyses within the VA informatics and Com- puting Infrastructure, VINCI. In both cases we used R version 4.0.3. Code used for analysis is available upon reasonable request and for collaborative purposes.2.6 Cox regression with inverse probability weighingWe hypothesized that elevated NLR is independently predictive of time-to-event of AD/ADRD and fit Cox PH models with inverse probability weighting (IPW). Time-to-event responses were calculated as the time from first NLR measurement (i.e., the origin) to the second occurrence of an AD/ADRD indicator (see definition provided). If two AD/ADRD ICD codes are documented in a single event, then the times- tamp of the single event is used. We used a cause-specific approach for Cox regression, treating deaths as censors. Follow-up was defined from baseline to the time of last encounter with the hospital system (NYU or VA), time of AD/ADRD diagnosis, or the date of a censoring event (death), whichever came first. Because of our interest in late-onset AD/ADRD, we excluded diagnoses that occurred before age 65. In observational and EHR studies, clinical and demographic covari- ates may affect both the NLR and risk of AD/ADRD (e.g., a confounding effect on both the predictor and the response) in the Cox analysis. As such, we weighted the survival model with the inverse of the fitted probabilities of dichotomized NLR levels from a logistic regres- sion model with age, sex, ethnicity (Hispanic), race, and histories of diabetes, hypertension, hypercholesteremia, CVD, and TBI. In both cohorts, weights were stabilized according to group prevalence 44 and truncated at the upper and lower 0.5% quantiles. We used Schoenfeld residuals (cox.zph function in R package survival) to validate the PH assumption. ### Page 5 HE ET AL . 5 of 11A natural spline with two degrees of freedom was used to model NLR, applying the same covariate adjustments and weighting scheme as described. To fully assess the contribution of non-linear effects toward AD/ADRD risk, variable importance was evaluated for the spline terms. This approach preserves the variation that would other- wise be lost through median dichotomization.2.7 Cumulative incidence function (CIF) analysisCIF curves were estimated with death as a competing risk event and for the dichotomized NLR. 9 We computed competing-risk CIF curves using R package causalCmprsk 45 with non-parametric subdistribution hazard model, and IPW to account for the confounding effects as in the primary Cox regression analysis.2.8 Subgroup analysisOwing to the large sample size we conducted separate analysis for four subgroups partitioned by age 65 and above at baseline time, sex, race, and ethnicity. Composite subgroups of these categories (race/ethnicity and sex stratified by age) were also considered. The same covariates were used as in the main analysis except in subgroups where they serve as the stratification criterion (e.g., race was omitted as a covariate for the Black/not Black subgroup). IPWs were recalculated within each subgroup to improve covariate balance in the corresponding subsam- ple, accounting for known differences in inflammatory profiles by age, sex, and race.2.9 Sensitivity analysesThe full potential observation window for VA patients spans from 2001 to 2023; however, we noted that inclusion of the full data induced temporal biases that incur violations of the Schoenfeld residuals at 0.01 significance for NLR values. For this reason, the primary anal- ysis for the VA cohort was restricted to those with baseline values (NLR sample dates) from 2011 to 2023 (same as in NYU), and we then compared blocks of contiguous years of data (2001–2013, 2006– 2018) with same lengths of 13 years and replicated the weighted Cox regressions from the primary analysis. In addition, we excluded outcomes occurring less than 6 months from the baseline time to reduce the risk of reverse causality. To further account for poten- tially differential long- and short-term effects of elevated NLR, either attributable to such reverse causality or an independent signal, 46 we used step-functions 47 to stratify the follow-up time at the first 3 months following the 6-month exclusion window of dementia inci- dence (for a total of 9 months past baseline). Follow-up times up to 9 months are categorized as short-term effects, whereas those later are considered long term. Finally, to evaluate potential bias related to NLR ascertainment setting and to assess the influence of transient elevations, we stratified analyses by whether the NLR mea- surement occurred within 3 days of an emergency department (ED) visit.3 RESULTS 3.1 Cohort characteristicsThe Inclusion diagram table for both cohorts is shown in Table S2. The summary of the demographic and clinical characteristics for all patients at baseline (time of NLR procurement) in the overall NYU and VA samples is shown grouped by median NLR is shown in Table 1. Dur- ing a window of 13 years from 2011 to 2023, the NYU cohort was composed of 284,530 patients with 4718 cases of incident AD/ADRD (1.7%) and 12,835 deaths (5%). The mean age was 69 (interquartile range [IQR] = 61, 75), and 55% of the cohort was female, with a mean follow-up time of 4.16 years (IQR = 1.44, 5.88) and a median NLR of 2.55 (IQR = 1.75, 4.14, mean [SD] = 3.52 [2.97]). The VA cohort com- prised 85,386 patients with 1753 cases of AD/ADRD (2%) and 17,759 deaths (21%). The mean age was 65 (IQR = 56, 70), and 6% of the cohort was female, with a mean follow-up time of 6.07 years (IQR = 3.02, 9.01) and the median NLR of 2.33 (IQR = 1.69, 3.27, mean [SD] = 2.73 [1.62]).3.2 Association of the neutrophil-to-lymphocyte ratio with AD/ADRD riskLogistic models used to estimate the IPW values indicated substan- tial differences for age, race, sex, and MH in both cohorts (Table S3). The association between nearly all covariates and elevated NLR was significant in both cohorts. Elevated NLR was positively and indepen- dently associated with the risk of AD/ADRD in both the NYU (hazard ratio [HR] 1.07, 95% CI = 1.02–1.15) and VA cohorts (HR = 1.21, 95% CI = 1.10–1.34). Spline fits for NLR showed that the partial coeffi- cients yield continuously positive trajectories across NLR ranges in both cohorts (Figure 1). Table 2 shows the HRs for all covariates in the adjusted Cox models. Distributional characteristics of NLR are shown in Figure S1. For both NYU and VA cohorts, histograms of weights after IPW estimation before and after truncation, as well as Love plots comparing standardized mean differences, are shown in Figure S2. CIF curves showed consistently higher incidence probabilities throughout the entire study period with mostly non-overlapping and parallel 95% CIs in the NYU and VA cohorts (Figure 2). Subgroup analyses by age, sex, race, and ethnicity showed consis- tent positive associations between NLR, with incident AD/ADRD for nearly all subgroups examined regardless of the study cohort (Table S4). In both cohorts, significantly higher HRs were observed for female (NYU HR = 1.05; 95% CI = 1.01–1.09), VA HR = 2.04; 95% CI = 1.24– 3.35) and Hispanic patients (NYU HR = 1.23; 95% CI = 1.00–1.51), VA HR = 1.42; 95% CI = 1.00–2.04). Ranked feature importance estimated from random forest survival models showed that spline terms of NLR had a moderately high pre- dictive value in the presence of the other covariates. The second term ### Page 6 6 of 11 HE ET AL .F I G U R E 1 Spline coefficients for the effects of increased NLR on cause-specific hazard of AD/ADRD (with 95% CI). Cause-specific hazard ratios of AD/ADRD incidence are shown corresponding to splines coefficients of NLR, scaled by standard deviation with 1 as the reference value. Spline terms for NLR are derived from cause-specific Cox models adjusted by age, sex, ethnicity (Hispanic), race, diabetes, hypertension, hypercholesteremia, CVD, and TBI. Cox models were inversely weighted to account for confounding effects of covariates. Inverse probability weights for the Cox regression were generated by the predicted probabilities of covariates on binarized NLR (1 if > median, 0 if < median) from a logistic regression model. AD, Alzheimer’s disease; ADRD, Alzheimer’s disease and related dementias; CI, confidence interval; CVD, cardio- and cerebrovascular disease; NLR, neutrophil-to-lymphocyte ratio; TBI, traumatic brain injury.F I G U R E 2 Adjusted CIF curves with 95% CIs for AD/ADRD for the NLR groups (binarized based on the median NLR). CIF curves were calculated from a non-parametric model using the causalCmprsk package. Higher NLR at baseline (date of NLR acquisition) was associated with greater probability of AD/ADRD in both cohorts. Models were adjusted for age, sex, race, ethnicity, hypertension, diabetes, history of CVD, and TBI. The observation window was from 2011 to 2023. AD, Alzheimer’s disease; ADRD, Alzheimer’s disease related dementias; CI, confidence interval; CIF, cumulative incidence function; CVD, cardio- and cerebrovascular disease; NLR, neutrophil-to-lymphocyte ratio; TBI, traumatic brain injury. ### Page 7 HE ET AL . 7 of 11TA B L E 1 Characteristics of NYU and VHA cohorts.NYU (median NLR = 2.55) VHA (median NLR = 2.33) Overall (n = 284,530 NLR < median (n = 142,551) NLR > median (n = 141,979)p-value (2-sided) Overall (n = 86,386) NLR < median (n = 42,824) NLR > median (n = 42,562)p-value (2-sided)Age, mean [Q1, Q3] 69 [61, 75] 69 [60, 75] 78 [72, 84] <0.001 65 [56, 70] 63 [56, 68] 66 [57, 73] <0.001 Female sex, n (%) 155,597 (55%) 85,011 (60%) 70,586 (50%) <0.001 7871 (9%) 5359 (13%) 2512 (5%) <0.001 Race/ethnicity White, n (%) 193,185 (68%) 89,721 (63%) 103,464 (73%) <0.001 63,716 (75%) 29,567 (69%) 34,149 (80%) <0.001 Black, n (%) 30,935 (11%) 20,288 (14%) 10,647 (7%) <0.001 11,362 (13%) 8006 (19%) 3356 (8%) <0.001 Asian, n (%) 25,315 (6%) 13,340 (6%) 11,975 (8%) <0.001 1658 (2%) 980 (2%) 678 (2%) <0.001 Other, n (%) 15,899 (6%) 8232 (6%) 7667 (5%) <0.001 7908 (9%) 3907 (9%) 4001 (9%) 0.17 Hispanic, n (%) 45,046 (16%) 24,640 (17%) 20,406 (14%) <0.001 5202 (6%) 2908 (7%) 2294 (6%) <0.001 Medical history (before baseline) Diabetes, n (%) 52,244 (18%) 26,365 (19%) 25,879 (18%) 0.07 17,402 (20%) 8464 (20%) 8938 (21%) <0.001 Hypertension, n (%) 123,750 (43%) 60,886 (43%) 62,864 (44%) <0.001 44,061 (52 %) 21,817 (51%) 22,244 (52%) <0.001 CVD, n (%) 113,044 (40%) 52,910 (37%) 60,134 (42%) <0.001 17,151 (20%) 7524 (18%) 9627 (23%) <0.001 TBI, n (%) 8938 (3%) 4182 (3%) 4756 (3%) <0.001 1460 (2%) 814 (2%) 646 (2%) <0.001 Hypercholest, n (%) 55,085 (19%) 30,301 (21%) 24,784 (17%) <0.001 11,529 (14%) 6106 (14%) 5423 (13%) <0.001 Outcome AD/ADRD, n (%) 4718 (1.7%) 2170 (1.5%) 2548 (1.8%) <0.001 1753 (2.0%) 703 (1.7%) 1050 (2.5%) <0.001 Deaths, n (%) 12,835 (5%) 4168 (3%) 8667 (6%) <0.001 17,759 (21%) 6223 (15%) 11,536 (27%) <0.001 Follow-up (years, mean [Q1, Q3]) 4.16 [1.44, 5.88] 4.53 [1.77, 6.75] 3.80 [1.17, 5.43] <0.001 6.07 [3.02, 9.01] 6.29 [3.26, 9.23] 5.86 [2.79, 8.73] <0.001 Exposure NLR, median [Q1, Q3] 2.55 [1.75, 4.14] 1.75 [1.36, 2.12] 4.04 [3.12, 6.17] <0.001 2.33 [1.69, 3.27] 1.69 [1.33, 2.00] 3.27 [2.73, 4.43] <0.001 NYU data were procured from all six NYU Langone Hospitals. VA data were extracted from the Corporate Data Warehouse that houses nationwide VA EHR. The p-values indicate two-sided t-tests for differences of across groups. Abbreviations: CVD, cardio- and cerebrovascular disease; NLR, neutrophil-to-lymphocyte ratio; NYU, New York University; TBI, traumatic brain injury; VA, Veterans Administration; VHA, Veteran’s Health Administration. appears immediately after age in NYU and both terms ranking after age and CVD in the VA, preceding the other covariates (Figure S3). Sensitivity analyses showed consistent NLR-ADRD associations across differing baseline time ranges in the VA cohort (Table S5). Anal- yses with time-stratified NLRs showed significant associations at both shorter (<9 months) and longer follow-up times but a pronounced decrease from shorter- to longer-term effect size in both cohorts (Table S6). Results stratified by NLR ascertainment setting (ED vs non-ED) show that the associations are driven primarily by non-ED exposures (Table S7).4 DISCUSSIONThis study supports the idea that peripheral inflammation and neutrophil biology contribute independently to the risk of future AD/ADRD. Systemic inflammation is important in neurodegenerative processes, and our results align with previous research that showed a link between inflammatory markers and cognitive decline. The study’s diverse sample, drawn from two major health care systems, reinforces the generalizability of these findings. Further research is needed to understand the biological mechanisms responsible for increases in NLR levels as well as the comorbidities, medication, and lifestyle factors influencing neutrophil inflammation in the context of AD/ADRD. The main advantage of this work is the inclusion and compari- son of two large, distinct EHR systems, which improve the external validity and generalizability of the findings. The sample cohorts from this study are much larger and extensive compared to those from prior research, 8,9,24 offering a clearer and more reliable assessment of the relationship between the NLR and risk of future AD/ADRD. Our results align with previous studies, 9 but they may be more reliable due to the larger sample sizes and more extensive adjustments. The adjusted CIFs in this analysis show clearer separation between high- and low-risk groups, unlike the overlapping intervals seen in earlier work. This provides stronger evidence of NLR’s role in AD/ADRD risk, accounting for the competing risk of death. Spline coefficients showed a monotonically increasing, almost linear, but distinctly inflected rela- tionship between NLR and AD/ADRD in both cohorts. In addition, the exploratory variable importance estimates suggest that NLR provides independent explanatory value for predicting AD/ADRD when consid- ### Page 8 8 of 11 HE ET AL .TA B L E 2 Full list of cause-specific adjusted hazard ratios for the association of the NLR and incident AD/ADRD in the NYU and VA cohorts.NYU (n = 284,530, events = 4718) VA (n = 85,836, events = 1753) aHR 95% CI aHR 95% CINLR (binary, above, or below median) 1.074 [1.022, 1.154] 1.211 [1.098, 1.337]Age 1.107 [1.103, 1.110] 1.111 [1.106, 1.117]Race (Asian) 0.740 [0.625, 0.877] 0.824 [0.484, 1.402] Race (Black) 1.029 [0.914, 1.156] 1.557 [1.323, 1.832]Race (Other) 1.048 [0.944, 1.163] 0.890 [0.736, 1.077] Female sex 1.000 [0.941, 1.064] 1.227 [0.963, 1.565] Hispanic ethnicity 1.221 [1.080, 1.380] 1.516 [1.255, 1.831]History of diabetes 1.325 [1.236, 1.420] 0.779 [0.688, 0.881]History of hypertension 1.093 [1.012, 1.181] 2.429 [2.117, 2.787]History of CVD 1.519 [1.405, 1.647] 2.951 [2.620, 3.324]History of hypercholesteremia 1.097 [1.025, 1.174] 0.733 [0.638, 0.841]History of TBI 1.859 [1.688, 2.048] 1.244 [0.802, 1.931] Adjusted hazard ratios (aHRs) from IPW Cox regressions for NYU and VHA cohorts. IPW Cox models were fit on the time scale for both cohorts. Significant associations at p < 0.05 highlighted in bold. Abbreviations: AD, Alzheimer’s disease; ADRD, Alzheimer’s disease and related dementias; CI, confidence interval; CVD, cardio- and cerebrovascular dis- ease; IPW, inverse probability weighting; NLR, neutrophil-to-lymphocyte ratio; NYU, New York University; TBI, traumatic brain injury; VA, Veterans Health Aministration. ered alongside other factors, supporting its potential as a biomarker, consistent with prior research. 9 Although these variable importance estimates supplement our primary findings and do not serve as defini- tive clinical evidence, they indicate that NLR retains predictive value even when modeled with established AD/ADRD risk factors. We implemented several measures to counteract potential biases. The NLR is known to be associated with AD/ADRD risk but is strongly influenced by comorbidities. 48 Because there is no random assign- ment to exposure groups (e.g., NLR levels), we controlled for these relationships by artificial median dichotomization with IPW of these exposures. 49 This separation offers two key advantages. First, median dichotomization reduces the small-scale variation in baseline NLR val- ues, which tend to be volatile and biologically noisy. By collapsing the variable into two levels, the Cox model and CIF analysis reduce potentially misleading variability seen in continuous NLR modeling. Dichotomization into high/low NLR levels simplifies interpretability and highlights the similarities in empirically observed thresholds that appear consistent across facilities (2.55 at NYU and 2.33 at the VA), which could inform future clinical decision rules. Second, dichotomiza- tion dovetails with IPW and CIF frameworks, which is more robust for binary exposures and required for valid CIF-based comparisons between groups. Alternative dichotomization methods to define high vs low NLR groups were also tested for sensitivity and results consis- tently showed positive HRs for higher NLR (not shown). To reduce the possibility of uncontrolled influences on inflamma- tion prior to baseline, including undocumented neurodegeneration, we additionally excluded outcomes between 0 and 6 months after base- line. We subset the data at baseline years 2011–2023, such that the two cohorts have matching study windows to reduce the variations associated with temporal misalignment. However, sensitivity analyses also revealed similarly significant effects of NLR in the VA across dif- fering baseline times. In addition, the lack of significant effects when limiting NLR measurements to those obtained in the ED indicates that the observed associations with AD/ADRD risk are not solely driven by transient elevations related to ED encounters. Although other sources of confounding in data collection may remain, these steps allowed us to rule out several major contributors. The differences in AD/ADRD risk across the two cohorts may be due to differences in sociodemographic characteristics or geography. Risk differences among subgroups are also likely attributable to these factors. The significantly higher HR among females in the VA may be related to the sex imbalance in the cohort (6% female compared to 55% in NYU), suggesting sex-specific burdens among veterans. NLR yields high HRs among the Hispanic subgroup in both the NYU and VA cohorts, although in the latter, the association is not significant (although nearly so) at the 95% level, perhaps due to the smaller sample size. These differences may be due to discrepancies in baseline inflam- mation, environmental stressors, or disparities in health care access, for example, geographic distributions for Hispanic populations in the New York Metropolitan Area. Future work may investigate the differ- ential avenues across cohorts in which these subgroups may influence the association of the NLR with risk for dementia. The increased risk of dementia associated with NLR values may be indicative of “inflammaging.” 20 Prior models that suggest that aging and inflammation lead to compromised neutrophil clearance may par- tially explain the relationships with amyloid and tau pathology. 50 Some younger subgroups in this study had a stronger risk signal with NLR, possibly because there are fewer confounding factors from other health conditions. But for the NYU and VA cohorts, the link between NLR and the risk of AD/ADRD for patients younger than 65 was ### Page 9 HE ET AL . 9 of 11more variable, with wider CIs, suggesting that age-modulated NLR risk toward ADRD may be complex and non-monotonic. A U-shaped pat- tern of NLR risk toward CVD has been noted before, 51 and future research could explore similar patterns in relation to dementia. To this end, some of the nonlinear relationships were captured using natural splines, followed by variable importance analysis to evaluate their contribution. These dose–response patterns are distinct from, but conceptually build upon, the primary analysis based on median- dichotomized NLR exposure. Although potential biases may arise from applying binary exposure-derived weights to these secondary analy- ses, these weightings are justifiable given the exploratory nature of this secondary analysis and the relative nascency of spline-based weighting methods. Limitations in the study of dementia through clinical labs arise from the problem of bidirectional causality: neurodegeneration also fuels inflammation through mechanisms such as sleep disturbance. 52,53 Neutrophils have a short lifespan 16 and its short-term role of inflamma- tion resolution through clearance 19 are much better understood than the long-term. 46 We aimed to study the long-term effects of NLR by excluding data from the 6 months before the ADRD outcome. Addi- tional Cox analyses showed that the HRs decreased from short to long term in both groups. This could be due to the lingering effects of prior neurodegeneration or the short-term impact of inflammation reflected in the NLR. Long-term effects were also significant and comparable with the overall HRs from the main analysis. We cannot fully deter- mine whether the short-term HR is attributable to reverse causality or an independent effect. However, the persistence of a significant asso- ciation in longer-term follow-up—after excluding outcomes occurring within the first 6 months after baseline and further separating early (6–9 months) from later follow-ups—supports our finding of consistent effects of NLR linked to chronic inflammation. Some limitations arise from the exclusion of covariates and observa- tions. We were not able to account for important AD risk factors such as genetic (e.g., APOE ε4 status), which has been reported to influence NLR levels, 24 or other variables associated with AD/ADRD risk such as years of education, cholesterol, history of psychiatric disease (e.g., depression), and triglyceride levels due to data availability, although prior studies suggest that the effects of these covariates are relatively small in the presence of more established factors. 9 In addition, com- plex interactions among MH variables (e.g., diabetes, hypertension, and CVD) may produce spurious associations. As such, only the NLR effect is directly interpretable in this study; future studies are needed to disentangle inflammation from downstream disease processes and related comorbidities.4.1 Further directionsWe evaluated associations of the NLR with subsequent AD/ADRD in the NYU and VA cohorts, and noted several differences, but investigat- ing the causes behind these discrepancies should be further explored in future work. Although not a diagnostic biomarker, our findings indicate that elevated NLR potentially yields additive value in dementia risk prediction models when combined with other key predictors. Future studies may explore its potential added value as a marker of disease progression in patients already identified to be at increased risk or at a preclinical stage in conjunction with AD molecular data (e.g., amyloid positivity). Exploratory variable importance analysis from this study suggests that NLR may be integrated into dementia prediction models to improve accuracy as a discrete indicator value (like in the Cox analy- sis of this study) or as a continuous value under either the A/T/N or I/V/S framework. Toward this end, future work may attempt to iden- tify potential clinical NLR thresholds that signal a differential risk for AD/ADRD specifically in the presence of specific comorbidities. Future studies should also consider longitudinal trajectories of NLR values in conjunction with other laboratory values reflecting inflammatory status.5 CONCLUSIONUsing data from two large, demographically and clinically unique clinical platforms, we confirmed that elevated levels of NLR are con- sistently associated with the risk of future AD/ADRD. In both datasets, we demonstrated that the association between elevated NLR and risk of AD/ADRD was generalizable and scalable across different set- tings and subgroups. Female and Hispanic participants consistently exhibited higher HRs for NLR in relation to the AD/ADRD risk, sug- gesting a potentially stronger association in these subgroups. Our study highlights the importance of peripheral inflammation and the need to further clarify the mechanisms by which neutrophil biology is implicated in AD/ADRD.ACKNOWLEDGMENTSThe study was supported by the National Institutes of Health (NIH; R01AG092953, R01AG070821, R01AG079282, and P30AG066512), National Alzheimer’s Coordinating Center AY2022/23 New Investiga- tor Award to J.R.C., and the Cooperative Studies Program, VA Boston Healthcare System. O.M.B. is supported by the NIH through the fol- lowing grants: K23AG068534, R01AG082278, and RF1AG083975; Alzheimer’s Association grant AARG-21-848397; and BrightFocus Foundation A2022033S.CONSENT STATEMENTHuman consent was not necessary for this study.CONFLICT OF INTEREST STATEMENTThe authors declare no conflicts of interest. Any author disclosures are available in the supporting information.ORCIDTianshe He https://orcid.org/0000-0001-9752-1256Jaime Ramos-Cejudo https://orcid.org/0000-0002-0993-9909 ### Page 10 10 of 11 HE ET AL .REFERENCES1. Jack CR, 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. 2. Jack CR, Bennett DA, Blennow K, et al. A/T/N: an unbiased descriptive classification scheme for Alzheimer disease biomarkers. Neurology. 2016;87(5):539-547. 3. Smyth LCD, Murray HC, Hill M, et al. Neutrophil-vascular interac- tions drive myeloperoxidase accumulation in the brain in Alzheimer’s disease. Acta Neuropathol Commun. 2022;10(1):38. 4. Kara S, Altunan B, Unal A. Investigation of the peripheral inflamma- tion (neutrophil-lymphocyte ratio) in two neurodegenerative diseases of the central nervous system. Neurol Sci. 2022;43(3):1799-1807. 5. Jacobs T, Jacobson SR, Fortea J, et al. The neutrophil to lympho- cyte ratio associates with markers of Alzheimer’s disease pathology in cognitively unimpaired elderly people. Immun Ageing. 2024;21(1):32. 6. Zenaro E, Pietronigro E, Bianca VD, et al. Neutrophils promote Alzheimer’s disease-like pathology and cognitive decline via LFA-1 integrin. Nat Med. 2015;21(8):880-886. 7. Luo J, Thomassen JQ, Nordestgaard BG, Tybjærg-Hansen A, Frikke- Schmidt R. Blood leukocyte counts in Alzheimer disease. JAMA Netw Open. 2022;5(10):e2235648. 8. Kuyumcu ME, Yesil Y, Oztürk ZA, et al. The evaluation of neutrophil- lymphocyte ratio in Alzheimer’s disease. Dement Geriatr Cogn Disord. 2012;34(2):69-74. 9. Ramos-Cejudo J, Johnson AD, Beiser A, et al. The neutrophil to lym- phocyte ratio is associated with the risk of subsequent dementia in the Framingham Heart Study. Front Aging Neurosci. 2021;13:773984. 10. Wang X, Wang B, Du X, et al. Associations between neutrophil- lymphocyte ratio and risk of cognitive impairment among Chinese older adults. BMC Geriatr. 2025;25:114. 11. Zhu Q, Zhang Q, Chen C, et al. C-reactive protein, neutrophil-to- lymphocyte ratio, and long-term mortality in Chinese centenarians.JAMA Netw Open. 2023;6(10):e2340307. 12. Serrano-López J, Martín-Antonio B. Inflammaging, an imbalanced immune response that needs to be restored for cancer prevention and treatment in the elderly. Cells. 2021;10(10):2562. 13. Roth-Walter F, Adcock IM, Benito-Villalvilla C, et al. Metabolic path- ways in immune senescence and inflammaging: novel therapeutic strategy for chronic inflammatory lung diseases. An EAACI posi- tion paper from the Task Force for Immunopharmacology. Allergy. 2024;79(5):1089-1122. 14. Liang X, Xiu C, Liu M, et al. Platelet-neutrophil interaction aggra- vates vascular inflammation and promotes the progression of atherosclerosis by activating the TLR4/NF-κB pathway. J Cell Biochem. 2019;120(4):5612-5619. 15. Dong Y, Lagarde J, Xicota L, et al. Neutrophil hyperactivation correlates with Alzheimer’s disease progression. Ann Neurol. 2018;83(2):387-405. 16. Pillay J, den Braber I, Vrisekoop N, et al. In vivo labeling with 2H2O reveals a human neutrophil lifespan of 5.4 days. Blood. 2010;116(4):625-627. 17. Kruger P, Saffarzadeh M, Weber ANR, et al. Neutrophils: between host defence, immune modulation, and tissue injury. PLoS Pathog. 2015;11(3):e1004651. 18. Ramos C, Oehler R, Clearance of apoptotic cells by neutrophils in inflammation and cancer. Cell Death Discov. 2024;10(1):26. 19. Bratton D, Henson P, Neutrophil clearance: when the party is over, clean-up begins. Trends Immunol. 2011;32(8):350-357. 20. Franceschi C, Bonafè M, Valensin S, et al. Inflamm-aging. An evo- lutionary perspective on immunosenescence. Ann N Y Acad Sci. 2000;908:244-254. 21. Luo J, Thomassen JQ, Nordestgaard BG, Tybjærg-Hansen A, Frikke- Schmidt R, Blood leukocyte counts in Alzheimer disease. JAMA Netw Open. 2022;5(10):e2235648x. 22. Baik SH, Cha M, Hyun Y, et al. Migration of neutrophils targeting amyloid plaques in Alzheimer’s disease mouse model. Neurobiol Aging. 2014;35(6):1286-1292. 23. Kalelioglu T, Yuruyen M, Gultekin G, et al. Neutrophil and platelet to lymphocyte ratios in people with subjective, mild cognitive impairment and early Alzheimer’s disease. Psychogeriatrics. 2017;17(6):506-508. 24. Rembach A, Watt AD, Wilson WJ, et al. An increased neutrophil- lymphocyte ratio in Alzheimer’s disease is a function of age and is weakly correlated with neocortical amyloid accumulation. J Neuroim- munol. 2014;273(1-2):65-71. 25. Chou OHI, Zhou J, Li L, et al. The association between neutrophil- lymphocyte ratio and variability with new-onset dementia: a population-based cohort study. J Alzheimers Dis. 2023;94(2):547-557. 26. Hou J, Ou Y, Xu W, Zhang P, Tan L, Yu J, Association of peripheral immunity with cognition, neuroimaging, and Alzheimer’s pathology.Alzheimers Res Ther. 2022;14(1):29. 27. Mehta NH, Zhou L, Li Y, et al. Peripheral immune cell imbalance is asso- ciated with cortical beta-amyloid deposition and longitudinal cognitive decline. Sci Rep. 2023;13(1):8847. 28. Rodríguez-Rodríguez E, López-Sobaler AM, Ortega RM, Delgado- Losada ML, López-Parra AM, Aparicio A, Association between neutrophil-to-lymphocyte ratio with abdominal obesity and healthy eating index in a representative older Spanish population. Nutrients. 2020;12(3):855. 29. Zhao T, Zhong T, Zhang M, Xu Y, Zhang M, Chen L, Alzheimer’s dis- ease: causal effect between obesity and APOE gene polymorphisms.Int J Mol Sci. 2023;24(17):13531. 30. Lee S, Byun MS, Yi D, et al. Body mass index and two-year change of in vivo Alzheimer’s disease pathologies in cognitively normal older adults. Alzheimers Res Ther. 2023;15(1):108. 31. Mertoglu C, Gunay M, Neutrophil-lymphocyte ratio and platelet- lymphocyte ratio as useful predictive markers of prediabetes and diabetes mellitus. Diabetes Metab Syndr. 2017;11(suppl 1):S127-S131. 32. Demirtas L, Degirmenci H, Akbas EM, et al. Association of hematolog- ical indices with diabetes, impaired glucose regulation and microvas- cular complications of diabetes. Int J Clin Exp Med. 2015;8(7):11420- 11427. 33. Lou M, Luo P, Tang R, et al. Relationship between neutrophil- lymphocyte ratio and insulin resistance in newly diagnosed type 2 diabetes mellitus patients. BMC Endocr Disord. 2015;15:9. 34. Luchsinger JA, Reitz C, Patel B, Tang M, Manly JJ, Mayeux R, Relation of diabetes to mild cognitive impairment. Arch Neurol. 2007;64(4):570- 575. 35. Roberts RO, Knopman DS, Geda YE, et al. Association of diabetes with amnestic and nonamnestic mild cognitive impairment. Alzheimers Dement. 2014;10(1):18-26. 36. Liu X, Zhang Q, Wu H, et al. Blood neutrophil to lymphocyte ratio as a predictor of hypertension. Am J Hypertens. 2015;28(11):1339-1346. 37. Jhuang Y, Kao T, Peng T, et al. Neutrophil to lymphocyte ratio as predictor for incident hypertension: a 9-year cohort study in Taiwan.Hypertens Res. 2019;42(8):1209-1214. 38. Glodzik L, Mosconi L, Tsui W, et al. Alzheimer’s disease markers, hyper- tension, and gray matter damage in normal elderly. Neurobiol Aging. 2012;33(7):1215-1227. 39. van Arendonk J, Neitzel J, Steketee RME, et al. Diabetes and hyper- tension are related to amyloid-beta burden in the population-based Rotterdam Study. Brain. 2023;146(1):337-348. 40. Launer LJ, Ross G, Petrovitch H, et al. Midlife blood pressure and dementia: the Honolulu-Asia aging study. Neurobiol Aging. 2000;21(1):49-55. 41. Price LE, Shea K, Gephart S, The Veterans Affairs Corporate Data Warehouse: uses and implications for nursing research and practice.Nurs Adm Q. 2015;39(4):311-318. 42. Ramos-Cejudo J, Corrigan JK, Zheng C, et al. Antidepressant exposure and long-term dementia risk in a nationwide retrospective study on US ### Page 11 HE ET AL . 11 of 11veterans with midlife major depressive disorder. Alzheimers Dement. 2024;20(6):4106-4114. 43. Logue MW, Dasgupta S, Farrer LA. Genetics of Alzheimer’s disease in the African American Population. J Clin Med. 2023;12(16):5189. 44. Xu S, Ross C, Raebel MA, Shetterly S, Blanchette C, Smith D, Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value Health. 2010;13(2):273- 277. 45. Vakulenko-Lagun B, Magdamo C, Charpignon ML, Zheng B, Albers MW, Das S, causalCmprsk: an R package for nonparametric and Cox- based estimation of average treatment effects in competing risks data.Comput Methods Programs Biomed. 2024;242:107819. 46. Herrero-Cervera A, Soehnlein O, Kenne E, Neutrophils in chronic inflammatory diseases. Cell Mol Immunol. 2022;19(2):177-191. 47. Li J, Modeling survival data: extending the Cox model. Sociol Methods Res. 2003;32(1):117-120. 48. Angkananard T, Anothaisintawee T, McEvoy M, Attia J, Thakkinstian A, Neutrophil lymphocyte ratio and cardiovascular disease risk: a system- atic review and meta-analysis. Biomed Res Int. 2018;2018:2703518. 49. Neumann A, Billionnet C, Covariate adjustment of cumulative inci- dence functions for competing risks data using inverse probability of treatment weighting. Comput Methods Programs Biomed. 2016;129:63- 70. 50. Cruz Hernández JC, Bracko O, Kersbergen CJ, et al. Neutrophil adhesion in brain capillaries reduces cortical blood flow and impairs memory function in Alzheimer’s disease mouse models. Nat Neurosci. 2019;22(3):413-420. 51. Li X, Liu M, Wang G, The neutrophil-lymphocyte ratio is associated with all-cause and cardiovascular mortality in cardiovascular patients.Sci Rep. 2024;14(1):26692. 52. Zhang W, Xiao D, Mao Q, Xia H, Role of neuroinflammation in neu- rodegeneration development. Signal Transduct Target Ther. 2023;8(1): 267. 53. Green TRF, Ortiz JB, Wonnacott S, Williams RJ, Rowe RK, The bidirectional relationship between sleep and inflammation links trau- matic brain injury and Alzheimer’s disease. Front Neurosci. 2020; 14:894.SUPPORTING INFORMATIONAdditional supporting information can be found online in the Support- ing Information section at the end of this article.How to cite this article: He T, Betensky RA, Osorio RS, et al. Neutrophil inflammation metrics are associated with the risk of future dementia in large data from NYU Langone Hospitals and the Veterans Health Administration. Alzheimer’s Dement. 2026;22:e71335. https://doi.org/10.1002/alz.71335