About

Richard J. Chappell is a Professor in the Department of Biostatistics and Medical Informatics at the University of Wisconsin–Madison. His research focuses on the analysis and design of clinical trials, link estimation in generalized linear models, bivariate and nonparametric survival analysis, predicting the course of Alzheimer’s Disease and prodromal AD, and models in radiobiology. He is involved in advancing methodologies related to biomedical informatics and biostatistics, contributing to the development of statistical approaches for medical research and clinical trial design.

Research topics

• Medicine
• Oncology
• Internal medicine
• Pathology
• Biology
• Sociology
• Cancer research
• Ecology
• Zoology
• Criminology
• Geography
• Radiology
• Demography
• Surgery
• Genetics
• Psychology

Selected publications

• Midlife sensory and motor measures among best predictors in parsimonious models of long‐term cognitive decline and incidence of cognitive impairment in aging adults

  Alzheimer s & Dementia · 2026-02-01

  articleOpen access

  INTRODUCTION: We aimed to construct a parsimonious risk prediction model of 10-year cognitive decline and impairment using factors measured in midlife. METHODS: Longitudinal data of N = 1,529 (mean age 49 years; 54% women) Beaver Dam Offspring Study participants were included. We assessed several health measures at baseline and 10-year cognitive decline and cognitive impairment. We constructed Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression models. RESULTS: The top cognitive decline predictors (age, income, fine-motor skills, olfaction, peripheral artery disease, and serum neurofilament light chain protein [NfL]) and cognitive impairment predictors (sex, fine-motor skills, olfaction, self-rated vision, alcohol consumption, and NfL) yielded areas under the receiver operating characteristic curves (AUCs) of 0.80 (95% confidence interval [0.76-0.83]) and 0.73[0.69-0.77], respectively. DISCUSSION: In middle-aged adults, sensory and motor function and NfL were among the best predictors of 10-year onset of cognitive decline and impairment. After external validation in other studies, these results could help identify those at risk for neurodegeneration and cognitive decline who could benefit from early interventions.

  PublisherDOI

• Community Counting Carnivores: Discrepancies between volunteer and DNR winter gray wolf counts in Wisconsin 2003-2011 

  2026-03-10

  articleOpen access

  Large carnivores are threatened globally, with most extant taxa having suffered significant historical range contractions. Due to this imperiled status, as well as increased scientific interest in top-down ecological processes, ecologists and conservationists have dedicated renewed efforts towards large carnivore preservation and management. As part of these efforts, reliable and transparent population monitoring is critical both to evaluating population dynamics as well as to detecting policy and management effects on large carnivore populations. Therefore, it is imperative to evaluate how methodological changes to monitoring regimes may affect the bias and uncertainty of estimates, especially with cryptic and politically contentious taxa like large carnivores. We describe methodological changes in Wisconsin gray wolf (Canis lupus) censusing techniques by the Wisconsin Department of Natural Resources (DNR), paying particular attention to a citizen science program where volunteers conducted winter wolf track surveys separately from DNR trackers. We hypothesize how changes to volunteer training and participation in winter wolf counts may have resulted in several methodologically distinct time series of wolf population estimates. To investigate this hypothesis, we use a Bayesian mixed effects model to analyze how volunteer and DNR trackers counted wolves during a relatively methodologically consistent period from 2003 to 2011 and find that volunteers counted 83% (95% CI: [74%-92%]) as many wolves as DNR trackers. Therefore, we conclude that changes in relative volunteer involvement before and after that period must necessarily affect the bias and precision of wolf population estimates. We hypothesize possible reasons for this discrepancy between volunteer and DNR trackers, including differences in tracking aptitude, potential biases among trackers, and differences in survey timing. We also simulate volunteer and DNR wolf counts as if both tracker types had surveyed all blocks across all years to compare our reproducible wolf count uncertainties to DNR-reported uncertainties. We end with recommendations for more transparent and reproducible wolf counting by the DNR and broader recommendations for ecological citizen science initiatives.

  PublisherDOI

• Ranked Set Sampling in Survival Analysis

  arXiv (Cornell University) · 2025-12-27

  preprintOpen accessSenior author

  Ranked set sampling (RSS) is a cost-efficient study design that uses inexpensive baseline ranking to select a more informative subset of individuals for full measurement. While RSS is well known to improve precision over simple random sampling (SRS) for uncensored outcomes, survival analysis under RSS has largely been limited to estimation of the Kaplan-Meier survival curve under random censoring. Consequently, many standard tools routinely used with SRS data, including log-rank and weighted log-rank tests, restricted mean survival time summaries, and window-based mean life measures, are not yet fully developed for RSS settings, particularly when ranking is imperfect and censoring is present. This work develops a unified survival analysis framework for balanced RSS designs that preserves efficiency gains while providing the inferential tools expected in applied practice. We formalize Kaplan-Meier and Nelson-Aalen estimators for right-censored data under both perfect and concomitant-based imperfect ranking and establish their large-sample properties using martingale and empirical process methods adapted to the rank-wise RSS structure. Rank-aware Greenwood-type variance estimators are proposed, and efficiency relative to SRS is evaluated through simulation studies varying set size, number of cycles, censoring proportion, and ranking quality. The framework is further extended to log-rank and Fleming-Harrington weighted tests, as well as restricted and window mean life functionals with asymptotic variance formulas and two-sample comparisons. An implementation plan with real-data illustrations is provided to facilitate practical use.

  PublisherDOI

• Assessing the effect of eicosapentaenoic acid (EPA) on CSF levels of matrix metalloproteinases and their inhibitors

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  Abstract Background Matrix metalloproteinases (MMPs) are a group of enzymes with roles in mediating extracellular matrix integrity and inflammation, and degrading beta amyloid fibrils. Studies have demonstrated that MMPs in atherosclerotic plaques can be downregulated by omega‐3 fatty acid eicosapentaenoic acid (EPA), but there has been little research looking at the effect of EPA on MMP levels in CSF. In this study, we evaluated the effect of an FDA approved high dose EPA supplement on CSF levels of MMP‐2 and MMP‐9, and MMP inhibitors TIMP‐1 and TIMP‐2. Method Cognitively healthy VA‐eligible veterans (ages 50‐76) were enrolled in the Brain Amyloid and Vascular Effects of Eicosapentaenoic Acid Study (BRAVE, NCT02719327), a randomized, placebo‐controlled, double‐blind, parallel‐group clinical trial with subject N = 128, randomized in a 1:1 ratio to receive either 4g icosapent ethyl (Vascepa® IPE) or a placebo (mineral oil) daily for 18 months. Participants with suspected memory impairment, liver or kidney disease, or those unable to comply with study procedures were excluded. CSF biomarkers MMP‐2, MMP‐9, TIMP‐1, and TIMP‐2 were obtained at baseline, 9 months, and 18 months. In these exploratory analyses, treatment group differences in CSF biomarkers (18 month/baseline ratio) were analyzed on an intent to treat basis using GraphPad Prism®. Result Contrary to published studies showing decreased levels and expression of MMP‐2 and ‐9 in individuals treated with EPA, our data revealed significant increases in levels of MMP‐2 ( p = 0.04) and MMP‐9 ( p = 0.003) in the IPE treated group compared to placebo. There was also a significant increase in TIMP‐2 in the IPE group ( p = 0.02), which is consistent with previously published in vitro studies. Although not statistically significant, the effect of IPE seems to be modified by APOE4 status, with a trend towards decreased MMP‐2, MMP‐9, and TIMP‐2 in APOE4‐positive individuals. Conclusion In this randomized clinical trial, veterans treated with high dose EPA (Vascepa® IPE) for 18 months demonstrated significant increases in CSF levels of MMP‐2, MMP‐9, and TIMP‐2 compared to individuals treated with placebo. These results contrast published studies showing decreased MMP levels in atherosclerotic plaques treated with EPA. Further research is needed to elucidate the effect of EPA on cerebral biomarkers.

  PublisherOA PDFDOI

• Factors influencing tau trajectories along the amyloid timeline from three cohorts

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Recent studies demonstrate tau burden is heterogeneous after A+ onset and is temporally proximal to clinical impairment in sporadic AD. This study uses temporal modeling and neuroimaging data from three cohorts to investigate common factors that may hasten amyloid-related tau accumulation. METHODS: Participants with available amyloid and tau PET imaging were included from ADNI (n = 880), OASIS (n = 445), and University of Wisconsin (WISC: WRAP and Wisconsin ADRC; n = 739) cohorts. The following steps were completed separately for each cohort. Amyloid and tau were quantified, respectively, using Centiloids (CL) and medial temporal and temporal neocortex standard uptake value ratios (SUVR). A+ and T+ thresholds were defined as the mean plus two standard deviations (SDs) of lower Gaussian mixture model distributions. Sampled iterative linear approximation (SILA) was used to estimate A+ onset age (EAOA) and A+ time (age at observation minus EAOA). To understand moderators of the relationship between A+ time and tau SUVR's, we excluded those deemed confidently A- (<1 SD below the lower GMM group mean) and used LMEs to characterize associations between A+ time and tau SUVR, and investigated whether age at tau baseline, APOE-e4 carriage, sex, or education category explained additional variation in tau, both as main effects and interactions with A+ time. RESULTS: Cohort characteristics are shown in Table 1. Results (Figure 1) were mostly consistent between OASIS and WISC cohorts with A+ time having a significant positive association with tau SUVR in both medial temporal and temporal neocortex, and APOE-e4 carriage having a significant interaction with A+ time for the medial temporal SUVR (APOE-e4 carriers had faster tau trajectories). These effects were also significant in ADNI, and additionally interactions of A+ time by baseline tau age and by sex (females had faster tau trajectories) were significant for medial temporal tau, and A+ time by baseline tau age for temporal neocortex (younger age had faster tau trajectories). The education by A+ time interaction did not reach significance in any cohort/region. CONCLUSION: In three longitudinal cohorts, APOE-e4 carriage consistently accelerated tau trajectories relative to A+ onset. Future work will further investigate these relationships and cohort differences that may contribute to mixed findings.

  PublisherOA PDFDOI

• Effects of Icosapent Ethyl, <i>APOE</i> Genotype and CSF Biomarkers on 4D Flow MRI Metrics of Vascular Health in Cognitively Unimpaired Veterans

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Veterans are more likely to develop vascular disease and Alzheimer's disease (AD) than non-veterans. Icosapent ethyl (IPE), a form of the omega-3 fatty acid eicosapentaenoic acid, reduces risk for major adverse cardiovascular events and lowers triglycerides. Improvement in vascular health may also reduce the risk of developing AD. 4D flow MRI, using PCVIPR (phase contrast vastly undersampled isotropic projection), offers direct study of blood flow and arterial stiffness in the brain. PCVIPR imaging may help uncover the effects of IPE on cerebrovascular health, and the relationship with APOE genotype and cerebrospinal fluid (CSF) biomarkers. METHODS: The Brain Amyloid and Vascular Effects of Eicosapentaenoic Acid (BRAVE-EPA) study enrolled VA-eligible, cognitively unimpaired Veterans, ages 50-75, at the Veteran's Hospital in Madison, Wisconsin. The study was a randomized, double-blind, placebo-controlled clinical trial of 4 g daily IPE (Vascepa®) treatment for 18 months. Participants had PCVIPR MRI, APOE genotyping and levels of CSF vascular health/inflammation biomarkers assayed with the NULISAseq CNS disease panel. PCVIPR imaging was completed on a 3T GE x750 scanner at baseline, month 9 and month 18. Image analysis was done with the Quantitative Velocity Tool in MATLAB. The PCVIPR measurements analyzed were mean blood flow (MF), pulsatility index (PI) and total cerebral blood flow (CBF; cervical internal carotid arteries + basilar artery). Interactions between variables were assessed with a Mann-Whitney U test and using correlation analysis. RESULTS: IPE treatment for 18 months did not significantly change MF, PI, or CBF measures (IPE: n = 41, placebo: n = 44; MF p-value range: 0.064-0.93; PI p-value range: 0.061-0.94; CBF p-value: 0.33). APOE ε4 carriers (n = 28) and non-carriers (n = 89) had similar baseline blood flow (MF p-value range: 0.16-0.92; PI p-value range: 0.055-0.98). In the full sample (n = 123), baseline CSF levels of vascular biomarkers were significantly correlated with baseline PI (p-value range: <0.0001-0.99; Figure 2). In contrast, few CSF vascular biomarkers were significantly correlated with MF (p-value range: 0.032-0.96; Figure 1). CONCLUSION: IPE treatment did not significantly impact PCVIPR measures of cerebral blood flow. Baseline vascular CSF biomarker levels, but not APOE ε4 carrier status, significantly correlated with baseline PCVIPR measures of PI and to a lesser extent MF.

  PublisherOA PDFDOI

• Effects of Icosapent Ethyl, <i>APOE</i> Genotype and CSF Biomarkers on 4D Flow MRI Metrics of Vascular Health in Cognitively Unimpaired Veterans

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  BACKGROUND: Veterans are more likely to develop vascular disease and Alzheimer's disease (AD) than non-veterans. Icosapent ethyl (IPE), a form of the omega-3 fatty acid eicosapentaenoic acid, reduces risk for major adverse cardiovascular events and lowers triglycerides. Improvement in vascular health may also reduce the risk of developing AD. 4D flow MRI, using PCVIPR (phase contrast vastly undersampled isotropic projection), offers direct study of blood flow and arterial stiffness in the brain. PCVIPR imaging may help uncover the effects of IPE on cerebrovascular health, and the relationship with APOE genotype and cerebrospinal fluid (CSF) biomarkers. METHODS: The Brain Amyloid and Vascular Effects of Eicosapentaenoic Acid (BRAVE-EPA) study enrolled VA-eligible, cognitively unimpaired Veterans, ages 50-75, at the Veteran's Hospital in Madison, Wisconsin. The study was a randomized, double-blind, placebo-controlled clinical trial of 4 g daily IPE (Vascepa®) treatment for 18 months. Participants had PCVIPR MRI, APOE genotyping and levels of CSF vascular health/inflammation biomarkers assayed with the NULISAseq CNS disease panel. PCVIPR imaging was completed on a 3T GE x750 scanner at baseline, month 9 and month 18. Image analysis was done with the Quantitative Velocity Tool in MATLAB. The PCVIPR measurements analyzed were mean blood flow (MF), pulsatility index (PI) and total cerebral blood flow (CBF; cervical internal carotid arteries + basilar artery). Interactions between variables were assessed with a Mann-Whitney U test and using correlation analysis. RESULTS: IPE treatment for 18 months did not significantly change MF, PI, or CBF measures (IPE: n = 41, placebo: n = 44; MF p-value range: 0.064-0.93; PI p-value range: 0.061-0.94; CBF p-value: 0.33). APOE ε4 carriers (n = 28) and non-carriers (n = 89) had similar baseline blood flow (MF p-value range: 0.16-0.92; PI p-value range: 0.055-0.98). In the full sample (n = 123), baseline CSF levels of vascular biomarkers were significantly correlated with baseline PI (p-value range: <0.0001-0.99; Figure 2). In contrast, few CSF vascular biomarkers were significantly correlated with MF (p-value range: 0.032-0.96; Figure 1). CONCLUSION: IPE treatment did not significantly impact PCVIPR measures of cerebral blood flow. Baseline vascular CSF biomarker levels, but not APOE ε4 carrier status, significantly correlated with baseline PCVIPR measures of PI and to a lesser extent MF.

  PublisherOA PDFDOI

• Accelerated Phenotypical Aging in Midlife is Associated With Long-Term Cognitive Decline

  Innovation in Aging · 2025-12-01

  articleOpen access

  Abstract PhenoAge, a multi-system biomarker uses easily-obtained common clinical blood tests and determines whether a person is younger or older on a biological and physiological level than expected by chronological age. Higher PhenoAge is associated with increased risk of disability, age-related morbidities and all-cause mortality. Its associations with early cognitive changes in midlife are less understood. The aim of this study was to determine whether accelerated PhenoAge in midlife was associated with 10-year cognitive changes in middle-aged to older adults. This longitudinal study is based on N = 2,630 (54% women;mean age 50years) Beaver Dam Offspring Study participants. We measured baseline blood-based clinical markers necessary for calculation of accelerated PhenoAge (PhenoAgeAccel). We tested Trail-making Test B (TMT-B) performance at baseline, 5-year and 10-year follow-up. We used linear mixed-effects model with PhenoAgeAccel as predictor and TMT-B time as outcome, adjusting for random intercepts, education and sex. Models were repeated sex-stratified. With every additional year older in PhenoAge compared to chronological age at baseline, participants performed worse on the TMT-B at baseline [overall:0.60 seconds slower, 95% Confidence Interval (0.29,0.91); women: 0.34(-0.05,0.72); men: 0.91(0.41,1.42)] and had a faster decline in TMT-B over 10-years [overall:0.94 seconds slower (0.58,1.29); women: 0.51(0.08,0.95); men: 1.48(0.90,2.07)]. Accelerated PhenoAge in midlife was associated with 10-year cognitive decline, overall and in men. Longer follow-up will be needed to determine whether PhenoAge might be predictive of cognitive impairment later in life. If confirmed, PhenoAge could become a cost-effective marker of cognitive decline and dementia and help identify at-risk individuals early and inform targeted interventions.

  PublisherOA PDFDOI

• Accelerated Phenotypical Aging in Midlife is Associated with Long‐Term Cognitive Decline in Middle‐Aged Adults

  Alzheimer s & Dementia · 2025-12-01

  articleOpen access

  Abstract Background PhenoAge is a multi‐system blood‐based aging marker that uses common clinical tests of glucose metabolism, inflammation and kidney and liver function. This easily‐obtained marker determines whether a person is younger or older on a biological and physiological level than expected by their chronological age. A higher PhenoAge is associated with increased risk of disability, age‐related morbidities and all‐cause mortality. Its associations with early cognitive changes in midlife is less understood. The aim of this study was to determine whether accelerated PhenoAge in midlife was associated with 10‐year cognitive changes in middle‐aged to older adults. Methods This longitudinal study is based on N = 2,630 (54% women; mean age 50 years;Table 1) Beaver Dam Offspring Study (BOSS) participants. We measured baseline blood‐based clinical markers of health necessary for calculation of PhenoAge and calculated accelerated PhenoAge (PhenoAgeAccel) as the residual of a linear model with chronological age as the predictor and PhenoAge as the outcome. We tested Trail‐making Test B (TMT‐B) performance at baseline, 5‐year and 10‐year follow‐up. We used a linear mixed‐effects model with PhenoAgeAccel as predictor and TMT‐B time as outcome, adjusting for random intercepts, sex and education. We repeated models stratified by sex. Results We found with every additional year older in PhenoAge compared to chronological age at baseline, participants performed worse on the TMT‐B at baseline [complete sample: 0.60 seconds slower, 95% Confidence Interval (0.29,0.91); women: 0.34 (‐0.05,0.72); men: 0.91 (0.41,1.42), Figure 1]. Moreover, with every additional year older in PhenoAge compared to chronological age at baseline, participants had a faster decline in TMT‐B over the 10‐year follow‐up [main effect + wave interaction: complete sample: 0.94 seconds slower (0.58,1.29); women: 0.51 (0.08,0.95); men: 1.48 (0.90,2.07);Figure 1]. Conclusion Accelerated PhenoAge in midlife was associated with cognitive decline over 10 years, overall and in men. Longer follow‐up will be needed to investigate sex differences further and to determine whether PhenoAge might be predictive of the onset of cognitive impairment later in life. If confirmed, PhenoAge could become a cost‐effective marker of cognitive decline and dementia and might help identify at‐risk individuals early. This could inform targeted prevention and treatment methods to promote healthy brain aging.

  PublisherOA PDFDOI

• Factors associated with age at tau pathology onset and time from tau onset to dementia

  medRxiv · 2025-03-13 · 2 citations

  preprintOpen access

  Abstract INTRODUCTION Elevated tau is temporally proximal to dementia onset but less is known about factors influencing T+ onset age and time to dementia following T+ in Alzheimer’s disease. We used sampled iterative localized approximation (SILA) estimated T+ onset age (ETOA) to investigate factors associated with T+ age and time from T+ to dementia onset in ADNI. METHODS Using SILA-estimated A+ and T+ onset ages derived from 18 F-Flortaucipir, 18 F-Florbetapir, and 18 F-Florbetaben PET and Cox proportional hazards and accelerated failure time models, we analyzed APOE , sex, amyloid burden, age, educational attainment, and literacy associations with ETOA and time from T+ to dementia. RESULTS Higher amyloid, APOE -ε4, lower education, and lower literacy associated with younger ETOA. Older ETOA and higher amyloid associated with shorter time from T+ to dementia. DISCUSSION This work highlights the prognostic value of ETOA and the need to better characterize factors contributing to ETOA and dementia onset in AD.

  PublisherOA PDFDOI

Recent grants

• Core C - Data Management and Statistical Core

  NIH · $55.0M · 2019–2029

• NIH Grant P50AG033514

  NIH · $31.8M · 2020

Frequent coauthors

• Wai Tong Ng

  University of Hong Kong - Shenzhen Hospital

  105 shared

• Cheuk‐Wai Choi

  University of Hong Kong

  97 shared

• Roger K.C. Ngan

  Chinese University of Hong Kong

  87 shared

• Brian O’Sullivan

  Princess Margaret Cancer Centre

  83 shared

• S. C. K. Law

  Queen Elizabeth Hospital

  81 shared

• Shui‐Yi Tung

  Mackay Memorial Hospital

  81 shared

• To‐Wai Leung

  Queen Mary Hospital

  81 shared

• K. K. Fung

  University of Wisconsin System

  81 shared

Education

• Ph.D., Biostatistics

  University of Wisconsin–Madison

  1990

• M.S., Biostatistics

  University of Wisconsin–Madison

  1986

• B.S., Mathematics

  University of Wisconsin–Madison

  1983