Distributional data analysis via quantile functions and its application to modeling digital biomarkers of gait in Alzheimer's Disease.

With the advent of continuous health monitoring with wearable devices, users now generate their unique streams of continuous data such as minute-level step counts or heartbeats. Summarizing these streams via scalar summaries often ignores the distributional nature of wearable data and almost unavoidably leads to the loss of critical information. We propose to capture the distributional nature of wearable data via user-specific quantile functions (QF) and use these QFs as predictors in scalar-on-quantile-function-regression (SOQFR). As an alternative approach, we also propose to represent QFs via user-specific L-moments, robust rank-based analogs of traditional moments, and use L-moments as predictors in SOQFR (SOQFR-L). These two approaches provide two mutually consistent interpretations: in terms of quantile levels by SOQFR and in terms of L-moments by SOQFR-L. We also demonstrate how to deal with multi-modal distributional data via Joint and Individual Variation Explained using L-moments. The proposed methods are illustrated in a study of association of digital gait biomarkers with cognitive function in Alzheimers disease. Our analysis shows that the proposed methods demonstrate higher predictive performance and attain much stronger associations with clinical cognitive scales compared to simple distributional summaries.

Wrist-Worn Accelerometry, Aging, and Gait Speed in the Baltimore Longitudinal Study of Aging.

Wrist-worn accelerometry metrics are not well defined in older adults. Accelerometry data from 720 participants (mean age 70 years, 55% women) were summarized into (a) total activity counts per day, (b) active minutes per day, (c) active bouts per day, and (d) activity fragmentation (the reciprocal of the mean active bout length). Linear regression and mixed-effects models were utilized to estimate associations between age and gait speed with wrist accelerometry. Activity counts per day, daily active minutes per day, and active bouts per day were negatively associated with age among all participants, while positive associations with activity fragmentation were only observed among those ≥65 years. More activity counts, more daily active minutes, and lower activity fragmentation were associated with faster gait speed. There were baseline age interactions with annual changes in total activity counts per day, active minutes per day, and activity fragmentation (Baseline age × Time, p < .01 for all). These results help define and characterize changes in wrist-based physical activity patterns among older adults.

Association between walking energy utilisation and longitudinal cognitive performance in older adults.

BACKGROUND: Human motor function is optimised for energetic efficiency, however, age-related neurodegenerative changes affects neuromotor control of walking. Energy utilisation has been associated with motor performance, but its association with cognitive performance is unknown. METHODS: The study population included 979 Baltimore Longitudinal Study of Aging participants aged $\ge$50 years (52% female, mean age: 70$\pm$10.2 years) with a median follow-up time of 4.7 years. Energy utilisation for walking was operationalised as a ratio of the energy cost of slow walking to peak walking energy expenditure during standardised tasks ('cost-ratio'). Cognitive functioning was measured using the Trail Making Tests, California Verbal Learning Test, Wechsler Adult Intelligence Scale (WAIS), letter and category fluency and card rotation tests. Linear mixed models adjusted for demographics, education and co-morbidities assessed the association between baseline cost-ratio and cognitive functioning, cross-sectionally and longitudinally. To investigate the relationship among those with less efficient energy utilisation, subgroup analyses were performed. RESULTS: In fully adjusted models, a higher cost-ratio was cross-sectionally associated with poorer performance on all cognitive tests except WAIS (P < 0.05 for all). Among those with compromised energy utilisation, the baseline cost-ratio was also associated with a faster decline in memory (long-delay free recall: ß = -0.4, 95% confidence interval [CI] = [-0.8, -0.02]; immediate word recall: ß = -1.3, 95% CI = [-2.7, 0.1]). CONCLUSIONS: These findings suggest cross-sectional and longitudinal links between energy utilisation and cognitive performance, highlighting an intriguing link between brain function and the energy needed for ambulation. Future research should examine this association earlier in the life course to gauge the potential for interventive mechanisms.

Registration of 24-hour accelerometric rest-activity profiles and its application to human chronotypes.

By collecting data continuously over 24 hours, accelerometers and other wearable devices can provide novel insights into circadian rhythms and their relationship to human health. Existing approaches for analyzing diurnal patterns using these data, including the cosinor model and functional principal components analysis, have revealed and quantified population-level diurnal patterns, but considerable subject-level variability remained uncaptured in features such as wake/sleep times and activity intensity. This remaining informative variability could provide a better understanding of chronotypes, or behavioral manifestations of one's underlying 24-hour rhythm. Curve registration, or alignment, is a technique in functional data analysis that separates "vertical" variability in activity intensity from "horizontal" variability in time-dependent markers like wake and sleep times; this data-driven approach is well-suited to studying chronotypes using accelerometer data. We develop a parametric registration framework for 24-hour accelerometric rest-activity profiles represented as dichotomized into epoch-level states of activity or rest. Specifically, we estimate subject-specific piecewise linear time-warping functions parametrized with a small set of parameters. We apply this method to data from the Baltimore Longitudinal Study of Aging and illustrate how estimated parameters give a more flexible quantification of chronotypes compared to traditional approaches.

Sociodemographic, Health and Lifestyle, Sampling, and Mental Health Determinants of 24-Hour Motor Activity Patterns: Observational Study.

BACKGROUND: Analyzing actigraphy data using standard circadian parametric models and aggregated nonparametric indices may obscure temporal information that may be a hallmark of the circadian impairment in psychiatric disorders. Functional data analysis (FDA) may overcome such limitations by fully exploiting the richness of actigraphy data and revealing important relationships with mental health outcomes. To our knowledge, no studies have extensively used FDA to study the relationship between sociodemographic, health and lifestyle, sampling, and psychiatric clinical characteristics and daily motor activity patterns assessed with actigraphy in a sample of individuals with and without depression/anxiety. OBJECTIVE: We aimed to study the association between daily motor activity patterns assessed via actigraphy and (1) sociodemographic, health and lifestyle, and sampling factors, and (2) psychiatric clinical characteristics (ie, presence and severity of depression/anxiety disorders). METHODS: We obtained 14-day continuous actigraphy data from 359 participants from the Netherlands Study of Depression and Anxiety with current (n=93), remitted (n=176), or no (n=90) depression/anxiety diagnosis, based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders, fourth edition. Associations between patterns of daily motor activity, quantified via functional principal component analysis (fPCA), and sociodemographic, health and lifestyle, sampling, and psychiatric clinical characteristics were assessed using generalized estimating equation regressions. For exploratory purposes, function-on-scalar regression (FoSR) was applied to quantify the time-varying association of sociodemographic, health and lifestyle, sampling, and psychiatric clinical characteristics on daily motor activity. RESULTS: Four components of daily activity patterns captured 77.4% of the variability in the data: overall daily activity level (fPCA1, 34.3% variability), early versus late morning activity (fPCA2, 16.5% variability), biphasic versus monophasic activity (fPCA3, 14.8% variability), and early versus late biphasic activity (fPCA4, 11.8% variability). A low overall daily activity level was associated with a number of sociodemographic, health and lifestyle, and psychopathology variables: older age (P<.001), higher education level (P=.005), higher BMI (P=.009), greater number of chronic diseases (P=.02), greater number of cigarettes smoked per day (P=.02), current depressive and/or anxiety disorders (P=.05), and greater severity of depressive symptoms (P<.001). A high overall daily activity level was associated with work/school days (P=.02) and summer (reference: winter; P=.03). Earlier morning activity was associated with older age (P=.02), having a partner (P=.009), work/school days (P<.001), and autumn and spring (reference: winter; P=.02 and P<.001, respectively). Monophasic activity was associated with older age (P=.005). Biphasic activity was associated with work/school days (P<.001) and summer (reference: winter; P<.001). Earlier biphasic activity was associated with older age (P=.005), work/school days (P<.001), and spring and summer (reference: winter; P<.001 and P=.005, respectively). In FoSR analyses, age, work/school days, and season were the main determinants having a time-varying association with daily motor activity (all P<.05). CONCLUSIONS: Features of daily motor activity extracted with fPCA reflect commonly studied factors such as the intensity of daily activity and preference for morningness/eveningness. The presence and severity of depression/anxiety disorders were found to be associated mainly with a lower overall activity pattern but not with the time of the activity. Age, work/school days, and season were the variables most strongly associated with patterns and time of activity, and thus future epidemiological studies on motor activity in depression/anxiety should take these variables into account.

Measuring circadian function in bipolar disorders: Empirical and conceptual review of physiological, actigraphic, and self-report approaches.

BACKGROUND: Interest in biological clock pathways in bipolar disorders (BD) continues to grow, but there has yet to be an audit of circadian measurement tools for use in BD research and practice. PROCEDURE: The International Society for Bipolar Disorders Chronobiology Task Force conducted a critical integrative review of circadian methods that have real-world applicability. Consensus discussion led to the selection of three domains to review-melatonin assessment, actigraphy, and self-report. RESULTS: Measurement approaches used to quantify circadian function in BD are described in sufficient detail for researchers and clinicians to make pragmatic decisions about their use. A novel integration of the measurement literature is offered in the form of a provisional taxonomy distinguishing between circadian measures (the instruments and methods used to quantify circadian function, such as dim light melatonin onset) and circadian constructs (the biobehavioral processes to be measured, such as circadian phase). CONCLUSIONS: Circadian variables are an important target of measurement in clinical practice and biomarker research. To improve reproducibility and clinical application of circadian constructs, an informed systematic approach to measurement is required. We trust that this review will decrease ambiguity in the literature and support theory-based consideration of measurement options.

Registration for exponential family functional data.

We introduce a novel method for separating amplitude and phase variability in exponential family functional data. Our method alternates between two steps: the first uses generalized functional principal components analysis to calculate template functions, and the second estimates smooth warping functions that map observed curves to templates. Existing approaches to registration have primarily focused on continuous functional observations, and the few approaches for discrete functional data require a pre-smoothing step; these methods are frequently computationally intensive. In contrast, we focus on the likelihood of the observed data and avoid the need for preprocessing, and we implement both steps of our algorithm in a computationally efficient way. Our motivation comes from the Baltimore Longitudinal Study on Aging, in which accelerometer data provides valuable insights into the timing of sedentary behavior. We analyze binary functional data with observations each minute over 24 hours for 592 participants, where values represent activity and inactivity. Diurnal patterns of activity are obscured due to misalignment in the original data but are clear after curves are aligned. Simulations designed to mimic the application indicate that the proposed methods outperform competing approaches in terms of estimation accuracy and computational efficiency. Code for our method and simulations is publicly available.

Contrasting characteristics of daily physical activity in older adults by cancer history.

BACKGROUND: Using objectively collected physical activity (PA) data from the Baltimore Longitudinal Study of Aging, the authors tested whether patterns of daily activity and sedentary time differed by cancer survivorship in older adults. METHODS: In total, 659 participants (mean age ± standard deviation, 71 ± 10 years; 51% women) who had self-reported information on cancer history were instructed to wear an accelerometer for 7 consecutive days. Accelerometer data were summarized into: 1) PA volume and 2) activity fragmentation (interrupted activity), expressed as both continuous and as dichotomized (low and high) variables. Participants were categorized into 4 groups by cross-classification of dichotomous PA volume and fragmentation. Multiple regression models were used to estimate differences in PA patterns by cancer history. RESULTS: Cancer survivors averaged 0.12 fewer log-transformed activity counts per day (standard error, 0.05; P = .02) than individuals who reported no history of cancer after adjusting for demographics, behavioral factors, and comorbidities. Although fragmentation did not differ by cancer survivorship in the continuous model (P = .13), cancer survivorship was associated with 77% greater odds (odds ratio, 1.77; 95% confidence interval, 1.11-2.82) of having high (vs low) fragmentation and 94% greater odds (odds ratio, 1.94; 95% confidence interval, 1.13-3.33) of having combined low PA/high fragmentation (vs high PA/low fragmentation) relative to those with no cancer history. CONCLUSIONS: The current findings suggest that cancer survivors engage in lower total daily PA and that they perform this activity in a more fragmented manner compared with adults without a history of cancer. These results may reflect the onset and progression of a low-activity phenotype that is more vulnerable to heightened levels of fatigue and functional decline with aging.

Fatigability and endurance performance in cancer survivors: Analyses from the Baltimore Longitudinal Study of Aging.

BACKGROUND: Fatigue is prevalent and distressing among cancer survivors, but its subjective nature makes it difficult to identify. Fatigability, defined as task-specific fatigue, and endurance performance may be useful supplemental measures of functional status in cancer survivors. METHODS: Fatigability, endurance performance, and cancer history were assessed every 2 years in Baltimore Longitudinal Study of Aging participants between 2007 and 2015. Fatigability was defined according to the Borg rating of perceived exertion scale after a 5-minute, slow treadmill walk; and endurance performance was calculated according to the ability and time to complete a fast-paced, 400-meter walk. The association between cancer history, fatigability, and endurance performance was evaluated using longitudinal analyses adjusted for age, sex, body mass index, and comorbidities. RESULTS: Of 1665 participants, 334 (20%) reported a history of cancer. A combination of older age (>65 years) and a history of cancer was associated with 3.8 and 8.6 greater odds of high perceived fatigability and poor endurance, respectively (P < .01). Older adults with and without a history of cancer walked 42 and 23 seconds slower than younger adults without a history of cancer, respectively (P < .01). The median times to the development of high fatigability and poor endurance were shorter among those who had a history of cancer compared with those who had no history of cancer (P < .01). CONCLUSIONS: The current findings suggest that a history of cancer is associated with fatigability and poor endurance and that this effect is significantly greater in older adults. Evaluating the effects of cancer and age on fatigability may illuminate potential pathways and targets for future interventions. Cancer 2018;124:1279-87. © 2018 American Cancer Society.

Two-way principal component analysis for matrix-variate data, with an application to functional magnetic resonance imaging data.

Many modern neuroimaging studies acquire large spatial images of the brain observed sequentially over time. Such data are often stored in the forms of matrices. To model these matrix-variate data we introduce a class of separable processes using explicit latent process modeling. To account for the size and two-way structure of the data, we extend principal component analysis to achieve dimensionality reduction at the individual level. We introduce necessary identifiability conditions for each model and develop scalable estimation procedures. The method is motivated by and applied to a functional magnetic resonance imaging study designed to analyze the relationship between pain and brain activity.

Novel metrics for growth model selection.

BACKGROUND: Literature surrounding the statistical modeling of childhood growth data involves a diverse set of potential models from which investigators can choose. However, the lack of a comprehensive framework for comparing non-nested models leads to difficulty in assessing model performance. This paper proposes a framework for comparing non-nested growth models using novel metrics of predictive accuracy based on modifications of the mean squared error criteria. METHODS: Three metrics were created: normalized, age-adjusted, and weighted mean squared error (MSE). Predictive performance metrics were used to compare linear mixed effects models and functional regression models. Prediction accuracy was assessed by partitioning the observed data into training and test datasets. This partitioning was constructed to assess prediction accuracy for backward (i.e., early growth), forward (i.e., late growth), in-range, and on new-individuals. Analyses were done with height measurements from 215 Peruvian children with data spanning from near birth to 2 years of age. RESULTS: Functional models outperformed linear mixed effects models in all scenarios tested. In particular, prediction errors for functional concurrent regression (FCR) and functional principal component analysis models were approximately 6% lower when compared to linear mixed effects models. When we weighted subject-specific MSEs according to subject-specific growth rates during infancy, we found that FCR was the best performer in all scenarios. CONCLUSION: With this novel approach, we can quantitatively compare non-nested models and weight subgroups of interest to select the best performing growth model for a particular application or problem at hand.

Re-evaluating the effect of age on physical activity over the lifespan.

Advancements in accelerometer analytic and visualization techniques allow researchers to more precisely identify and compare critical periods of physical activity (PA) decline by age across the lifespan, and describe how daily PA patterns may vary across age groups. We used accelerometer data from the 2003-2006 cohorts of the National Health and Nutrition Examination Survey (NHANES) (n=12,529) to quantify total PA as well as PA by intensity across the lifespan using sex-stratified, age specific percentile curves constructed using generalized additive models. We additionally estimated minute-to-minute diurnal PA using smoothed bivariate surfaces. We found that from childhood to adolescence (ages 6-19) across sex, PA is sharply lower by age partially due to a later initiation of morning PA. Total PA levels, at age 19 are comparable to levels at age 60. Contrary to prior evidence, during young adulthood (ages 20-30) total and light intensity PA increases by age and then stabilizes during midlife (ages 31-59) partially due to an earlier initiation of morning PA. We additionally found that males compared to females have an earlier lowering in PA by age at midlife and lower total PA, higher sedentary behavior, and lower light intensity PA in older adulthood; these trends seem to be driven by lower PA in the afternoon compared to females. Our results suggest a re-evaluation of how emerging adulthood may affect PA levels and the importance of considering time of day and sex differences when developing PA interventions.

Quantifying the lifetime circadian rhythm of physical activity: a covariate-dependent functional approach.

Objective measurement of physical activity using wearable devices such as accelerometers may provide tantalizing new insights into the association between activity and health outcomes. Accelerometers can record quasi-continuous activity information for many days and for hundreds of individuals. For example, in the Baltimore Longitudinal Study on Aging physical activity was recorded every minute for [Formula: see text] adults for an average of [Formula: see text] days per adult. An important scientific problem is to separate and quantify the systematic and random circadian patterns of physical activity as functions of time of day, age, and gender. To capture the systematic circadian pattern, we introduce a practical bivariate smoother and two crucial innovations: (i) estimating the smoothing parameter using leave-one-subject-out cross validation to account for within-subject correlation and (ii) introducing fast computational techniques that overcome problems both with the size of the data and with the cross-validation approach to smoothing. The age-dependent random patterns are analyzed by a new functional principal component analysis that incorporates both covariate dependence and multilevel structure. For the analysis, we propose a practical and very fast trivariate spline smoother to estimate covariate-dependent covariances and their spectra. Results reveal several interesting, previously unknown, circadian patterns associated with human aging and gender.

Normalization and extraction of interpretable metrics from raw accelerometry data.

We introduce an explicit set of metrics for human activity based on high-density acceleration recordings from a hip-worn tri-axial accelerometer. These metrics are based on two concepts: (i) Time Active, a measure of the length of time when activity is distinguishable from rest and (ii) AI, a measure of the relative amplitude of activity relative to rest. All measurements are normalized (have the same interpretation across subjects and days), easy to explain and implement, and reproducible across platforms and software implementations. Metrics were validated by visual inspection of results and quantitative in-lab replication studies, and by an association study with health outcomes.

Generalized multilevel function-on-scalar regression and principal component analysis.

This manuscript considers regression models for generalized, multilevel functional responses: functions are generalized in that they follow an exponential family distribution and multilevel in that they are clustered within groups or subjects. This data structure is increasingly common across scientific domains and is exemplified by our motivating example, in which binary curves indicating physical activity or inactivity are observed for nearly 600 subjects over 5 days. We use a generalized linear model to incorporate scalar covariates into the mean structure, and decompose subject-specific and subject-day-specific deviations using multilevel functional principal components analysis. Thus, functional fixed effects are estimated while accounting for within-function and within-subject correlations, and major directions of variability within and between subjects are identified. Fixed effect coefficient functions and principal component basis functions are estimated using penalized splines; model parameters are estimated in a Bayesian framework using Stan, a programming language that implements a Hamiltonian Monte Carlo sampler. Simulations designed to mimic the application have good estimation and inferential properties with reasonable computation times for moderate datasets, in both cross-sectional and multilevel scenarios; code is publicly available. In the application we identify effects of age and BMI on the time-specific change in probability of being active over a 24-hour period; in addition, the principal components analysis identifies the patterns of activity that distinguish subjects and days within subjects.

Structured functional principal component analysis.

Motivated by modern observational studies, we introduce a class of functional models that expand nested and crossed designs. These models account for the natural inheritance of the correlation structures from sampling designs in studies where the fundamental unit is a function or image. Inference is based on functional quadratics and their relationship with the underlying covariance structure of the latent processes. A computationally fast and scalable estimation procedure is developed for high-dimensional data. Methods are used in applications including high-frequency accelerometer data for daily activity, pitch linguistic data for phonetic analysis, and EEG data for studying electrical brain activity during sleep.

MRI Evaluation of Radiofrequency, Cryothermal, and Laser Left Atrial Lesion Formation in Patients with Atrial Fibrillation.

BACKGROUND:  Catheter ablation utilizing radiofrequency (RF), Cryothermal (Cryo), or Laser energy is effective for treatment of atrial fibrillation (AF). Late gadolinium enhancement magnetic resonance imaging (LGE-MRI) has been used to estimate the burden of left atrial (LA) fibrosis, but no data exist regarding structural changes following each modality. We sought to compare the baseline to postprocedure change in LA scar burden following RF, Cryo, or Laser ablation for treatment of AF. METHODS: Seventeen patients with AF underwent initial pulmonary vein (PV) isolation (PVI) using RF (n = 7), Cryo (n = 5), and Laser (n = 5). LGE-MRI was performed prior to and at 24 hours and 3 months after PVI. RESULTS: In a linear mixed-effects model, accounting for intrapatient clustering of data and interpatient differences in baseline scar, LGE extent was significantly increased at 24 hours postablation (+14.6 ± 1.9% of LA myocardium, P < 0.001), and remained stable from 24 hours to 3 months (+0.12 ± 1.9%, P = 0.951). There was no statistically significant difference between the postablation scar extent among ablation modalities when compared to RF (Cryo +4.5 ± 3.0%, P = 0.123; Laser -3.2 ± 3.0%, P = 0.291). The PV antral LGE intensity was increased by 25.1 ± 3.8% (P<0.001) 24 hours after ablation and additionally increased by 8.1 ± 3.8 at 3 months (P = 0.033). CONCLUSIONS: Radiofrequency, Cryo, and laser ablation result in increased LGE extent and intensity at 24 hours and 3 months postablation. No statistically significant difference was noted in the extent of fibrosis induced by any modality.

Single nucleotide polymorphisms in proximity to K-channel genes are associated with decreased longitudinal QTc variance.

BACKGROUND: Atrial fibrillation (AF) is the most common form of cardiac arrhythmia. Despite significant progress in identification of predisposing factors, the pathophysiology of AF remains to be elucidated. Previous studies have reported that single nucleotide polymorphisms (SNPs) in potassium-channel genes associate with AF and the instantaneous corrected QT interval (QTc). The purpose of this study was to examine the association between SNPs in proximity to KCNQ1, KCNH2, KCNE2, and KCNJ2 and longitudinal QTc variations in patients with AF. METHODS AND RESULTS: We conducted a retrospective cohort study of 800 electrocardiograms from 93 patients with AF. All patients were Caucasian, with an average age of 61.2 years, and 72% were male. Of all patients, 37% had persistent AF, and 63% had paroxysmal AF. Following DNA extraction from blood, SNPs at the AF-associated loci KCNQ1, KCNH2, KCNE2, and KCNJ2 were genotyped using the Sequenom MassARRAY. Using a linear regression model and adapting a resampling inference, a decrease in longitudinal QTc variance was found to associate with SNPs near KCNH2 (rs10240738) and KCNJ2 (rs8079702) when adjusted for patient age, gender, AF type, and average QTc. On average, patients with these SNPs had a shorter QTc interval. In addition, we fitted a multilevel mixed effects regression model accounting for subject level heterogeneity and found no longitudinal association between presence of SNPs near K-channel genes and changes in QTc. CONCLUSION: Polymorphisms near specific potassium-channel genes in AF patients are associated with decreased longitudinal QTc variance and a shorter average QTc. These results support the hypothesis that effects on myocardial repolarization may mediate the association of these SNPs and AF.

Quantifying the time-varying association between objectively measured physical activity and mortality in US older adults over a 12-year follow-up period: the NHANES 2003-2006 study.

Objectively measuring physical activity (PA) has consistently shown an association with reduced all-cause mortality risk in cross-sectional studies. However, the strength of this association may change over time. We quantify the time-varying, covariate-adjusted association between the total volume of PA and all-cause mortality over a 12-year follow-up period using Cox regression with a time varying effect of population-referenced quantile total activity count adjusted for traditional risk factors. Analyses focus on participants 50-84 years old with adequate accelerometer wear time and without missing covariates. The findings suggest that (1) the use of baseline PA in Cox models with long follow-up periods may be inappropriate without time-varying effects and (2) the use of accelerometry derived volume of PA in risk score calculations may be most appropriate for short-term to medium-term risk scores.

Actigraphy Estimated Sleep Moderates the Relationship between Physical Activity and Cognition in Older Adults.

Background and Aims: Physical inactivity and poor sleep are common in older adults and may interact to contribute to age- and disease-related cognitive decline. However, prior work regarding the associations among physical activity, and cognition in older adults is primarily limited to subjective questionnaires that are susceptible to inaccuracies and recall bias. Therefore, this study examined whether objectively measured physical activity and sleep characteristics, each estimated using actigraphy, are independently or interactively associated with cognitive performance. Methods: The study included 157 older adults free of dementia (136 cognitively unimpaired; 21 MCI; M age = 71.7) from the BIOCARD cohort. Results: Using multiple linear regression, cognition was regressed on estimated total volume of physical activity (TVPA), sleep efficiency (SE), wake after sleep onset (WASO), and total sleep time (TST) (adjusted for age, sex, education, diagnosis, vascular risk factors, and Apolipoprotein E (APOE)-e4 genetic status). Models were also run for domain-specific cognitive composite scores. TVPA and SE each were positively associated with a global cognitive composite score. TVPA was positively associated with executive function and language composites, and SE was positively related to executive function, visuospatial, and language composites. Importantly, a TVPA by SE interaction (p = 0.015) suggested that adults with the poorest SE experienced the greatest benefit from physical activity in relation to global cognition. The other sleep metrics were unrelated to cognitive performance. Conclusion: These results suggest that TVPA and SE may synergistically benefit cognition in older adults.