About

Rebecca A Betensky is the Chair of the Department of Biostatistics and a Professor of Biostatistics at NYU School of Global Public Health. Prior to her current position, she was a Professor of Biostatistics at the Harvard T.H. Chan School of Public Health, where she also served as director of the Harvard Catalyst Biostatistics Program, the Data and Statistics Core for the Massachusetts Alzheimer’s Disease Research Center, and the Biostatistics Neurology Core at Massachusetts General Hospital. Her research focuses on methods for the analysis of censored and truncated outcomes and covariates, which frequently arise from the subsampling of cohort studies. She has a long-standing interest in clinical trials, particularly in the evaluation of biomarkers and the use and interpretation of p-values. Dr. Betensky has collaborated extensively on studies in neurologic diseases and serves as the statistical editor for the Annals of Neurology. She has directed NIH training programs in neurostatistics and neuroepidemiology, and has been involved in promoting diversity within biostatistics. Her professional service includes membership on NIH study sections, committees for the Institute of Medicine, and advisory roles for transplant registries. She is an elected Fellow of the American Statistical Association and the International Statistical Institute, and a past recipient of the Spiegelman Award from the American Public Health Association. Currently, she serves on the Board of Scientific Counselors for Clinical Science and Epidemiology at the National Cancer Institute.

Research topics

• Medicine
• Internal medicine
• Biology
• Oncology
• Demography
• Pharmacology
• Radiology
• Endocrinology
• Emergency medicine
• Gerontology
• World Wide Web
• Pathology
• Environmental health
• Cancer research
• Pediatrics
• Bioinformatics

Selected publications

• 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 & Dementia · 2026-04-01

  articleOpen access

  INTRODUCTION: Neutrophil-to-lymphocyte ratio (NLR), a marker of systemic inflammation, 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; 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.

  PublisherDOI

• Test-retest reliability of FreeSurfer measures of neurodegeneration

  NeuroImage · 2026-04-10

  articleOpen access

  Reliable structural brain measurements are essential for studying neurodegeneration and for designing adequately powered aging and Alzheimer's disease (AD) research. We evaluated the test-retest reliability of FreeSurfer 7.1 morphometric measures in 100 older adults (mean age 73.5 years) ranging from cognitively unimpaired to dementia. Each participant underwent two T1-weighted 3T MRI scans on the same scanner within a short interval (mean 5.5 weeks), minimizing biological change. Segmentation was performed in both standard cross-sectional and longitudinal FreeSurfer modes, focusing on AD-relevant volumes of entorhinal cortex, hippocampus, lateral ventricles, choroid plexus, and the AD cortical thickness signature. Reliability was quantified using absolute and root-mean-square test-retest differences, standard deviation of differences, and intraclass correlation coefficients. Longitudinal processing improved precision by 15-50% across most measures compared with cross-sectional processing, with the largest gain observed for entorhinal thickness. Larger, anatomically well-defined regions (e.g., hippocampus, AD signature) demonstrated higher reliability than small structures or those with complex geometry (e.g., entorhinal cortex, choroid plexus). Image quality, indexed by the Euler characteristic, was the only factor significantly associated with measurement variability; reliability was unrelated to age, sex, cognitive status, inter-scan interval, or amyloid/tau PET burden. Power analyses indicated that detecting a 1% within-individual change requires sample sizes ranging from 36 (AD signature) to >300 (entorhinal cortex). We observed low reliability of choroid plexus volumetry by FreeSurfer 7. These results provide practical benchmarks for expected FreeSurfer measurement variability in older adults. They highlight the advantages of longitudinal processing and rigorous quality control for research on brain aging and AD.

  PublisherDOI

• Combining p‐tau217 and digital cognitive testing to predict cognitive decline

  Alzheimer s & Dementia · 2026-03-30

  articleOpen access

  INTRODUCTION: We investigated whether a multi-day learning curve (MDLC), derived from remote digital cognitive assessments over 1 week, adds prognostic value to plasma phosphorylated tau at threonine 217 (p-tau217) for predicting cognitive decline in cognitively unimpaired (CU) older adults. METHODS: A total of 183 CU participants (mean age 73.3; 70% female) completed multi-day digital cognitive tests and provided plasma p-tau217. Linear mixed-effects models evaluated the independent and combined effects of baseline p-tau217 and MDLC on longitudinal cognitive decline over 2.5 years. RESULTS: Adding MDLC to p-tau217 significantly explained longitudinal cognitive trajectory beyond p-tau217 alone. Although higher p-tau217 and lower MDLC were independently associated with steeper decline, individuals with both risk factors declined the fastest. Notably, those with elevated p-tau217 level but preserved MDLC remained stable, distinguishing them from those at imminent risk. DISCUSSION: Combining digital learning metrics with plasma p-tau217 enables precise risk stratification. This approach identifies individuals most vulnerable to imminent decline, potentially enhancing screening for early preventive interventions.

  PublisherDOI

• Association of Social Determinants of Health with Brain MRI Outcomes in Individuals with Pediatric Onset Multiple Sclerosis (S3.003)

  Neurology · 2025-04-07

  article

  To investigate the association between childhood social determinants of health (SDOH) and brain MRI outcomes in individuals with pediatric onset multiple sclerosis (POMS).

  PublisherDOI

• Neutrophil inflammation metrics are associated with the risk of future dementia in large‐scale electronic health record data from hospital systems

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Neutrophils play a role in Alzheimer's disease (AD) pathology and AD-related dementias (AD/ADRD), and prior research has shown that the neutrophil to lymphocyte ratio (NLR), a marker of neutrophil-mediated inflammation, is associated with the risk of future dementia. To date, studies looking at this relationship have used small cohorts. We address whether this is generalizable to larger populations using electronic health records (EHR) data from all six sites of NYU Langone Hospitals METHOD: Our study window ranged from 2011-2023. NLR values were obtained from laboratory test results. For each patient, the first NLR obtained was used; the associated date was taken to be the time origin. The outcome was AD/ADRD incidence, defined using ICD-codes over the same study window at least 6 months post-baseline. Cause-specific Cox regression was used to determine the independent association of log-transformed NLR values with the risk of future AD/ADRD adjusting for demographic and clinical confounders and with death as a censoring event. To account for confounding effects of covariates, the Cox model was weighted by the predicted probabilities of all adjusting covariates on binarized NLR ("high" if >median, "low" if <median) from a logistic model. We used cumulative incidence function (CIF) curves to assess the risk over time stratified by binarized NLR accounting for the competing risk of death. Sex- and race- subgroup analysis was also conducted. RESULT: The study sample included 287,985 patients at NYU (4,773 AD/ADRD cases, mean age [Q1, Q3] = 69 [61, 75], 55% female, 11% black/AA). We found a positive and independent association of log NLR with cause-specific hazard of AD/ADRD (HR = 1.160 with 95% CI [1.100, 1.223]. Spline terms for log NLR indicated significant increasing relationships with cause-specific hazard of AD/ADRD. Subgroup analyses showed consistently significant associations, with higher hazard among Hispanic patients. CIF curves showed meaningful separation for high vs low-NLR groups. CONCLUSION: Our findings suggest that individuals with higher NLR are at greater risk of future AD/ADRD regardless of the study cohort or demographics, reinforcing the importance of peripheral inflammation and the need to clarify the mechanisms by which neutrophil biology influences AD pathology.

  PublisherOA PDFDOI

• Long‐COVID‐19 Symptoms are associated with Abnormal Cognitive Testing and Mild Cognitive Impairment Compared to Controls

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: There is a paucity of data comparing neuropsychiatric testing in patients with post-acute COVID-19 symptoms ("long-COVID") to COVID-19 patients without long-term symptoms and to COVID-19 negative controls. METHOD: We conducted a cross-sectional study of patients with and without prior laboratory-confirmed COVID-19 (COV+ vs COV-), who had no previous history of dementia/cognitive impairment. Long-COVID was defined as subjective symptoms (WHO symptom questionnaire) lasting >1 month after index SARS-CoV-2 infection. All patients underwent UDS3 neuropsychiatric testing and received a formal cognitive diagnosis (with respective etiology following NACC criteria) by a consensus group of physicians blinded to their long-COVID status as: normal, mild-cognitive impairment (MCI), or dementia. Cognitive test scores, CDR, and physician diagnoses were compared between patients with and without long-COVID using Fisher's exact and Mann-Whitney U tests and multivariable logistic regression models were constructed to adjust for differences in age, sex, time from initial COVID diagnosis and number of COVID-19 infections. RESULT: We enrolled 279 patients: N = 51 (18%) COV-, N = 228 (82%) COV+, N = 122 (44%) with long-COVID, median age 70 years (IQR 63-76), and 62% female. Compared to COV- subjects and COV+ subjects without long-COVID, participants with long-COVID had significantly worse cognitive test results (≥1 cognitive battery abnormal in 64% with long-COVID compared to 46% without, p = 0.003), as well as worse global CDR and CDR sum of boxes scores (26% CDR>0 among long-COVID vs. 6% without long-COVID, p <0.001). Significantly more long-COVID patients were diagnosed with MCI or dementia (25% vs. 6% without long-COVID, p <0.001), including higher proportions with MCI related to Alzheimer's disease (10% vs. 3%, p = 0.038), and MCI related to psychiatric diagnoses (11% vs. 1%, p <0.001). After adjusting for sex, age, time from initial COVID diagnosis and number of COVID-19 infections, long-COVID was associated with significantly higher odds of an MCI diagnosis (aOR 3.5, 95% CI 1.2-10.6, p = 0.025), and specifically MCI due to Alzheimer's (aOR 4.4, 95% CI 1.1-18.3, p = 0.041), but not MCI due to psychiatric disease. CONCLUSION: Long-COVID patients had significantly worse cognitive test scores and higher rates of MCI (Alzheimer's type) diagnoses compared to both COV- and COV+ subjects without long-COVID symptoms.

  PublisherOA PDFDOI

• Alcohol Sales and Adverse Events Due to the Covid-19 Pandemic — A Natural Experiment

  NEJM Evidence · 2025-02-25

  editorial1st authorCorresponding
  PublisherDOI

• Markers of Blood‐Brain Barrier Permeability and Immune Exhaustion are Present in Patients with Post‐COVID‐19 Brain Fog Compared to Controls

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: While mechanisms underlying post-COVID-19 cognitive impairment are unknown, increased blood-brain barrier (BBB) permeability and immune alterations have been posited as possible contributing factors. METHODS: We conducted a cross-sectional study of patients with and without prior laboratory-confirmed COVID-19 (COV+ vs COV-), with no previous history of dementia/cognitive impairment. Brain fog was defined as the subjective symptoms of memory loss, confusion, and/or difficulty concentrating lasting >1 month after index SARS-CoV-2 infection. Fasting plasma biomarkers of inflammation (cytokines), and BBB disruption (Simoa SP-X), as well as phosphorylated tau (pTau) were measured using Simoa HD-X technology. Patients underwent concurrent UDS3 neuropsychiatric testing and received a formal cognitive diagnosis by a consensus group of physicians following NACC criteria. Plasma biomarker levels were compared between patients with and without brain fog, and with and without a diagnosis of MCI using Mann-Whitney U tests. RESULTS: We enrolled 279 patients: N = 51 (18%) COV-, N = 228 (82%) COV+, N = 96 (34%) with post-COVID brain fog, and N = 35 (13%) COV+ patients with newly diagnosed MCI post-COVID. Several plasma cytokine levels (TNF-a, IL-4, IL-10, IL-22) were significantly lower in those with brain fog compared to those without (all p ≤0.01), while markers of BBB permeability were higher (heparin-binding epidermal growth factor 10.1 pg/mL in brain fog patients vs 7.7 pg/mL without, p <0.001; vascular endothelial growth factor 72 vs. 63 pg/mL, p = 0.055; and placental growth factor 12.4 vs 11.4 pg/mL, p = 0.090). Among COV+ patients diagnosed with MCI, TNF-a was significantly lower compared to cognitively normal subjects (1.3 vs. 1.5 pg/mL, p = 0.031), while pTau-217 was significantly higher in those with MCI due to Alzheimer's (0.21 vs. 0.13 pg/mL, p = 0.009) compared to those without. CONCLUSIONS: Lower cytokine levels, increased markers of BBB disruption, and altered markers of tau pathology were observed in patients with brain fog and COV+ subjects with MCI, which may suggest immune exhaustion, increased BBB permeability, and disrupted tau processing as possible mechanisms of post-COVID-19 brain fog.

  PublisherOA PDFDOI

• Body mass index and blood volume influence plasma biomarkers and positron emission tomography classification in preclinical Alzheimer's disease

  Alzheimer s & Dementia · 2025-10-01 · 8 citations

  articleOpen access

  Abstract INTRODUCTION Blood‐based biomarkers (BBMs) are promising tools for Alzheimer's disease (AD) diagnosis, but their accuracy may be affected by body mass index (BMI) and blood volume (BV) through dilution. We investigated how BMI and BV influence BBM concentrations and PET prediction. METHODS Data from 241 cognitively unimpaired participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI) were examined to evaluate the influence of BMI/BV on BBMs (Aβ 42/40 , p‐Tau 181 , p‐Tau 217 , glial fibrillary acidic protein [GFAP], neurofilament light chain [NfL]) and BBM‐based PET predictions. RESULTS Elevated BMI/BV associated with lower BBM concentrations, especially for p‐Tau 217 and NfL, independent of brain amyloid burden. BMI‐stratified thresholds improved amyloid PET prediction, with higher BBM thresholds and area under the curve (AUC) values seen in normal weight compared to overweight or obese participants. Drastic BMI/BV declines due to weight loss increased BBM variability and systematic PET misclassification. DISCUSSION Adjusting for BMI/BV in BBM‐based diagnostics appears to improve accuracy and reliable detection of AD pathology, especially in preclinical stages. Highlights Body mass index (BMI) and blood volume (BV) significantly influenced plasma BBM concentrations in cognitively unimpaired (CU) individuals. Blood‐based biomarkers (BBMs) associated more strongly with BV than with BMI. Dilution effects were independent of brain amyloid burden. BMI‐stratified BBM thresholds improved amyloid positron emission tomography (PET) classification accuracy. Declines in BMI/BV resulted in PET prediction bias and systematic errors.

  PublisherOA PDFDOI

• Cox Regression on the Plane

  ArXiv.org · 2025-09-15

  preprintOpen access

  The Cox proportional hazards model is the most widely used regression model in univariate survival analysis. Extensions of the Cox model to bivariate survival data, however, remain scarce. We propose two novel extensions based on a Lehmann-type representation of the survival function. The first, the simple Lehmann model, is a direct extension that retains a straightforward structure. The second, the generalized Lehmann model, allows greater flexibility by incorporating three distinct regression parameters and includes the simple Lehmann model as a special case. For both models, we derive the corresponding regression formulations for the three bivariate hazard functions and discuss their interpretation and model validity. To estimate the regression parameters, we adopt a bivariate pseudo-observations approach. For the generalized Lehmann model, we extend this approach to accommodate a trivariate structure: trivariate pseudo-observations and a trivariate link function. We then propose a two-step estimation procedure, where the marginal regression parameters are estimated in the first step, and the remaining parameters are estimated in the second step. Finally, we establish the consistency and asymptotic normality of the resulting estimators. We illustrate the approach using data from the Global Retinoblastoma Outcome Study.

  PublisherOA PDFDOI

Recent grants

• NIH Grant R01CA075971

  NIH · $1.3M · 2013

• NIH Grant R03CA121884

  NIH · $162k · 2010

• Statistical methods for censored and dependently truncated data

  NIH · $1.3M · 2018–2023

• NIH Grant R03CA105956

  NIH · $164k · 2006

• NIH Grant T32NS048005

  NIH · $3.0M · 2019

Frequent coauthors

• Tracy T. Batchelor

  Dana-Farber Brigham Cancer Center

  297 shared

• Joachim M. Baehring

  Yale University

  251 shared

• Jean‐Yves Blay

  Centre Léon Bérard

  245 shared

• Brian Patrick O’Neill

  Henry Ford Health System

  245 shared

• Jennie Taylor

  University of California, San Francisco

  200 shared

• Marc C. Chamberlain

  nLIGHT (United States)

  196 shared

• Khê Hoang‐Xuan

  Institut du Cerveau

  196 shared

• Antonio Omuro

  196 shared

Awards & honors

• Elected Fellow of the American Statistical Association
• Elected Fellow of the International Statistical Institute
• Spiegelman Award from the American Public Health Association