Bayesian models for multiple outcomes nested in domains

Bayesian outcome selection modeling

In psychiatric and social epidemiology studies, it is common to measure multiple different outcomes using a comprehensive battery of tests thought to be related to an underlying construct of interest. In the research that motivates our work, researchers wanted to assess the impact of in utero alcohol exposure on child cognition and neuropsychological development, which are evaluated using a range of different psychometric tests. Statistical analysis of the resulting multiple outcomes data can be challenging, because the outcomes measured on the same individual are not independent. Moreover, it is unclear, a priori, which outcomes are impacted by the exposure under study. While researchers will typically have some hypotheses about which outcomes are important, a framework is needed to help identify outcomes that are sensitive to the exposure and to quantify the associated treatment or exposure effects of interest. We propose such a framework using a modification of stochastic search variable selection, a popular Bayesian variable selection model and use it to quantify an overall effect of the exposure on the affected outcomes. The performance of the method is investigated empirically and an illustration is given through application using data from our motivating study.

Discovering structure in multiple outcomes models for tests of childhood neurodevelopment

Bayesian model-based clustering provides a powerful and flexible tool that can be incorporated into regression models to better understand the grouping of observations. Using data from the Seychelles Child Development Study, we explore the effect of prenatal methylmercury exposure on 20 neurodevelopmental outcomes measured in 9-year-old children. Rather than cluster individual subjects, we cluster the outcomes within a multiple outcomes model. By using information in the data to nest the outcomes into groups called domains, the model more accurately reflects the shared characteristics of neurodevelopmental domains and improves estimation of the overall and outcome-specific exposure effects by shrinking effects within and between domains selected by the data. The Bayesian paradigm allows for sampling from the posterior distribution of the grouping parameters; thus, inference can be made about group membership and their defining characteristics. We avoid the often difficult and highly subjective requirement of a priori identification of the total number of groups by incorporating a Dirichlet process prior to form a fully Bayesian multiple outcomes model.

Time and frequency dependent changes in resting state EEG functional connectivity following lipopolysaccharide challenge in rats

Research has shown that inflammatory processes affect brain function and behavior through several neuroimmune pathways. However, high order brain functions affected by inflammation largely remain to be defined. Resting state functional connectivity of synchronized oscillatory activity is a valid approach to understand network processing and high order brain function under different experimental conditions. In the present study multi-electrode EEG recording in awake, freely moving rats was used to study resting state connectivity after administration of lipopolysaccharides (LPS). Male Wistar rats were implanted with 10 cortical surface electrodes and administered with LPS (2 mg/kg) and monitored for symptoms of sickness at 3, 6 and 24 h. Resting state connectivity and power were computed at baseline, 6 and 24 h. Three prominent connectivity bands were identified using a method resistant to spurious correlation: alpha (5–15 Hz), beta-gamma (20–80 Hz), and high frequency oscillation (150–200 Hz). The most prominent connectivity band, alpha, was strongly reduced 6 h after LPS administration, and returned to baseline at 24 h. Beta-gamma connectivity was also reduced at 6 h and remained reduced at 24 h. Interestingly, high frequency oscillation connectivity remained unchanged at 6 h and was impaired 24 h after LPS challenge. Expected elevations in delta and theta power were observed at 6 h after LPS administration, when behavioral symptoms of sickness were maximal. Notably, gamma and high frequency power were reduced 6 h after LPS and returned to baseline by 24 h, when the effects on connectivity were more evident. Finally, increases in cross-frequency coupling elicited by LPS were detected at 6 h for theta-gamma and at 24 h for theta-high frequency oscillations. These studies show that LPS challenge profoundly affects EEG connectivity across all identified bands in a time-dependent manner indicating that inflammatory processes disrupt both bottom-up and top-down communication across the cortex during the peak and resolution of inflammation.

USING THE SEYCHELLES CHILD DEVELOPMENT STUDY TO CLUSTER MULTIPLE OUTCOMES INTO DOMAINS TO IMPROVE ESTIMATION OF THE OVERALL EFFECT OF MERCURY ON NEURODEVELOPMENT

Environmental exposure effects on human development can be small and difficult to detect due to the nature of observational data. In the Seychelles Child Development Study, researchers examined the effect of prenatal methylmercury exposure using a battery of tests measuring aspects of child development [23, 25]. We build a multiple outcomes model similar to that of the previous analyses (see [23, 25]); however, our multiple outcomes model makes no assumptions of relationships between the testing outcomes. Instead, the nesting of outcomes into domains is a clustering problem we address with a Dirichlet process mixture model implemented through a Bayesian MCMC approach [16]. This model provides inference for the methylmercury exposure effect as well as greater insight into the similarities and differences across the outcomes.

Estimating scaled treatment effects with multiple outcomes

In classical study designs, the aim is often to learn about the effects of a treatment or intervention on a single outcome; in many modern studies, however, data on multiple outcomes are collected and it is of interest to explore effects on multiple outcomes simultaneously. Such designs can be particularly useful in patient-centered research, where different outcomes might be more or less important to different patients. In this paper, we propose scaled effect measures (via potential outcomes) that translate effects on multiple outcomes to a common scale, using mean-variance and median-interquartile range based standardizations. We present efficient, nonparametric, doubly robust methods for estimating these scaled effects (and weighted average summary measures), and for testing the null hypothesis that treatment affects all outcomes equally. We also discuss methods for exploring how treatment effects depend on covariates (i.e., effect modification). In addition to describing efficiency theory for our estimands and the asymptotic behavior of our estimators, we illustrate the methods in a simulation study and a data analysis. Importantly, and in contrast to much of the literature concerning effects on multiple outcomes, our methods are nonparametric and can be used not only in randomized trials to yield increased efficiency, but also in observational studies with high-dimensional covariates to reduce confounding bias.

Association of Stressful Life Events with Psychological Problems: A Large-Scale Community-Based Study Using Grouped Outcomes Latent Factor Regression with Latent Predictors

Objective

The current study is aimed at investigating the association between stressful life events and psychological problems in a large sample of Iranian adults.

Method

In a cross-sectional large-scale community-based study, 4763 Iranian adults, living in Isfahan, Iran, were investigated. Grouped outcomes latent factor regression on latent predictors was used for modeling the association of psychological problems (depression, anxiety, and psychological distress), measured by Hospital Anxiety and Depression Scale (HADS) and General Health Questionnaire (GHQ-12), as the grouped outcomes, and stressful life events, measured by a self-administered stressful life events (SLEs) questionnaire, as the latent predictors.

Results

The results showed that the personal stressors domain has significant positive association with psychological distress (β = 0.19), anxiety (β = 0.25), depression (β = 0.15), and their collective profile score (β = 0.20), with greater associations in females (β = 0.28) than in males (β = 0.13) (all P < 0.001). In addition, in the adjusted models, the regression coefficients for the association of social stressors domain and psychological problems profile score were 0.37, 0.35, and 0.46 in total sample, males, and females, respectively (P < 0.001).

Conclusion

Results of our study indicated that different stressors, particularly those socioeconomic related, have an effective impact on psychological problems. It is important to consider the social and cultural background of a population for managing the stressors as an effective approach for preventing and reducing the destructive burden of psychological problems.

Impact of Mild Head Injury on Neuropsychological Performance in Healthy Older Adults: Longitudinal Assessment in the AIBL Cohort

Traumatic brain injury (TBI) is suggested to be a significant risk factor for dementia. However, little research has been conducted into long-term neuropsychological outcomes after head trauma. Participants from the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL) who had recovered after sustaining a mild TBI involving loss of consciousness more than 5 years previously were compared with matched controls across a 3-year period. Bayesian nested-domain modeling was used to estimate the effect of TBI on neuropsychological performance. There was no evidence for a chronic effect of mild TBI on any neuropsychological domain compared to controls. Within the TBI group, there was some evidence suggesting that the age that the head trauma occurred and the duration of unconsciousness were modulators of episodic memory. However, these findings were not robust. Taken together, these findings indicate that adults who have sustained a TBI resulting in loss of consciousness, but who recover to a healthy level of cognitive functioning, do not experience frank deficits in cognitive ability.

The effects of dexamphetamine on the resting-state electroencephalogram and functional connectivity

The catecholamines-dopamine and noradrenaline-play important roles in directing and guiding behavior. Disorders of these systems, particularly within the dopamine system, are associated with several severe and chronically disabling psychiatric and neurological disorders. We used the recently published group independent components analysis (ICA) procedure outlined by Chen et al. (2013) to present the first pharmaco-EEG ICA analysis of the resting-state EEG in healthy participants administered 0.45 mg/kg dexamphetamine. Twenty-eight healthy participants between 18 and 41 were recruited. Bayesian nested-domain models that explicitly account for spatial and functional relationships were used to contrast placebo and dexamphetamine on component spectral power and several connectivity metrics. Dexamphetamine led to reductions across delta, theta, and alpha spectral power bands that were predominantly localized to Frontal and Central regions. Beta 1 and beta 2 power were reduced by dexamphetamine at Frontal ICs, while beta 2 and gamma power was enhanced by dexamphetamine in posterior regions, including the parietal, occipital-temporal, and occipital regions. Power-power coupling under dexamphetamine was similar for both states, resembling the eyes open condition under placebo. However, orthogonalized measures of power coupling and phase coupling did not show the same effect of dexamphetamine as power-power coupling. We discuss the alterations of low- and high-frequency EEG power in response to dexamphetamine within the context of disorders of dopamine regulation, in particular schizophrenia, as well as in the context of a recently hypothesized association between low-frequency power and aspects of anhedonia. Hum Brain Mapp 37:570-588, 2016. © 2015 Wiley Periodicals, Inc.

Hierarchical regression for analyses of multiple outcomes

In cohort mortality studies, there often is interest in associations between an exposure of primary interest and mortality due to a range of different causes. A standard approach to such analyses involves fitting a separate regression model for each type of outcome. However, the statistical precision of some estimated associations may be poor because of sparse data. In this paper, we describe a hierarchical regression model for estimation of parameters describing outcome-specific relative rate functions and associated credible intervals. The proposed model uses background stratification to provide flexible control for the outcome-specific associations of potential confounders, and it employs a hierarchical "shrinkage" approach to stabilize estimates of an exposure's associations with mortality due to different causes of death. The approach is illustrated in analyses of cancer mortality in 2 cohorts: a cohort of dioxin-exposed US chemical workers and a cohort of radiation-exposed Japanese atomic bomb survivors. Compared with standard regression estimates of associations, hierarchical regression yielded estimates with improved precision that tended to have less extreme values. The hierarchical regression approach also allowed the fitting of models with effect-measure modification. The proposed hierarchical approach can yield estimates of association that are more precise than conventional estimates when one wishes to estimate associations with multiple outcomes.

Longitudinal cognitive decline in the AIBL cohort: The role of APOE ε4 status

The ε4 polymorphism of the APOE gene confers a substantially increased risk of developing Alzheimer's disease. However, the influence of the ε4 allele on age-related cognitive functioning is more contentious. Previously, we demonstrated relatively little evidence for a role of the ε4 allele on baseline cognitive performance in older adults in the Australian Imaging, Biomarkers and Lifestyle (AIBL) Study of Ageing (Foster et al., 2013). We here investigated whether the APOE ε4 allele influenced cognitive status over time when the AIBL cohort was studied longitudinally over a 3-year period. The AIBL neuropsychological test battery was administered at baseline, after 18 months and again after 36 months. Participants comprised 764 Healthy Controls and 131 Mild Cognitively Impaired individuals enrolled in the AIBL Study of Ageing. We compared individuals within each group with and without an ε4 allele. Healthy Controls with an ε4 allele manifested a modest acceleration in cognitive decline over 36 months on measures of verbal episodic memory. By contrast, Mild Cognitively Impaired individuals with an ε4 allele showed increased cognitive decline across a range of cognitive tasks, putatively reflecting early cognitive signs of Alzheimer's disease. Given the long prodromal period that has been noted in late onset Alzheimer's disease, we suggest that these findings are consistent with a prodromal account rather than a phenotypic account of ε4-related cognitive ageing.

Neuropsychological performance is positively associated with plasma albumin in healthy adults

BACKGROUND

Albumin serves a range of physiological functions that are vital to overall brain and cognitive health. Indeed, associations between cognitive performance and albumin have been demonstrated in individuals with chronic liver or kidney disease and in patients with a high urinary excretion of albumin. However, an association of plasma albumin with cognitive performance has not been reported in otherwise healthy participants with clinically acceptable plasma albumin concentrations.

METHOD

This study utilized a wide-ranging neuropsychological test battery to investigate the relationship between cognitive performance and plasma albumin homeostasis in 222 healthy participants (143 females) between the ages of 43 and 84 years (mean 65 years).

RESULTS

Albumin both with and without the covariates of age, sex and acute-phase proteins was positively associated with enhanced performance on a range of neuropsychological domains including perceptual speed, Stroop and verbal ability. Albumin manifested generally positive but less robust associations with secondary and primary memory.

CONCLUSION

The results indicate that there is a positive association between albumin and cognitive performance in physiologically healthy participants free of chronic renal or liver disease.

Bayesian Models for Multiple Outcomes in Domains with Application to the Seychelles Child Development Study

The Seychelles Child Development Study (SCDS) examines the effects of prenatal exposure to methylmercury on the functioning of the central nervous system. The SCDS data include 20 outcomes measured on 9-year old children that can be classified broadly in four outcome classes or "domains": cognition, memory, motor, and social behavior. Previous analyses and scientific theory suggest that these outcomes may belong to more than one of these domains, rather than only a single domain as is frequently assumed for modeling. We present a framework for examining the effects of exposure and other covariates when the outcomes may each belong to more than one domain and where we also want to learn about the assignment of outcomes to domains. Each domain is defined by a sentinel outcome which is preassigned to that domain only. All other outcomes can belong to multiple domains and are not preassigned. Our model allows exposure and covariate effects to differ across domains and across outcomes within domains, and includes random subject-specific effects which model correlations between outcomes within and across domains. We take a Bayesian MCMC approach. Results from the Seychelles study and from extensive simulations show that our model can effectively determine sparse domain assignment, and at the same time give increased power to detect overall, domain-specific and outcome-specific exposure and covariate effects relative to separate models for each endpoint. When fit to the Seychelles data, several outcomes were classified as partly belonging to domains other than their originally assigned domains. In retrospect, the new partial domain assignments are reasonable and, as we discuss, suggest important scientific insights about the nature of the outcomes. Checks of model misspecification were improved relative to a model that assumes each outcome is in a single domain.

The Serum Concentration of the Calcium Binding Protein S100B is Positively Associated with Cognitive Performance in Older Adults

S100B is a calcium binding peptide produced predominantly by astroglial cells in the central nervous system. S100B paradoxically has neurotrophic and apoptotic effects, dependent on extracellular concentration. This study investigated the relationship between serum S100B levels and neuropsychological performance across a range of cognitive domains in healthy older aged adults. A cohort of 219 participants between the ages of 43 and 84 years (141 female) were recruited. Subjects provided a fasting blood sample for S100B measurement (Mean = 0.24 ng/mL, SD = 0.14) and completed a battery of neuropsychological tests. S100B concentrations (both with and without the covariates of age and sex) were positively associated with the following measures of cognitive performance: digit-symbol coding, Stroop test, and measures of verbal ability. The results from this study show that serum S100B is positively associated with better cognitive performance in healthy older adults.

Latent factor regression models for grouped outcomes

We consider regression models for multiple correlated outcomes, where the outcomes are nested in domains. We show that random effect models for this nested situation fit into a standard factor model framework, which leads us to view the modeling options as a spectrum between parsimonious random effect multiple outcomes models and more general continuous latent factor models. We introduce a set of identifiable models along this spectrum that extend an existing random effect model for multiple outcomes nested in domains. We characterize the tradeoffs between parsimony and flexibility in this set of models, applying them to both simulated data and data relating sexually dimorphic traits in male infants to explanatory variables.

Alternative methods for testing treatment effects on the basis of multiple outcomes: simulation and case study

In clinical trials multiple outcomes are often used to assess treatment interventions. This paper presents an evaluation of likelihood-based methods for jointly testing treatment effects in clinical trials with multiple continuous outcomes. Specifically, we compare the power of joint tests of treatment effects obtained from joint models for the multiple outcomes with univariate tests based on modeling the outcomes separately. We also consider the power and bias of tests when data are missing, a common feature of many trials, especially in psychiatry. Our results suggest that joint tests capitalize on the correlation of multiple outcomes and are more powerful than standard univariate methods, especially when outcomes are missing completely at random. When outcomes are missing at random, test procedures based on correctly specified joint models are unbiased, while standard univariate procedures are not. Results of a simulation study are reported, and the methods are illustrated in an example from the Clinical Antipsychotic Trials of Intervention Effectiveness for schizophrenia.

Interpreting epidemiological evidence in the presence of multiple endpoints: an alternative analytic approach using the 9-year follow-up of the Seychelles child development study

PURPOSE

The potential for ill-informed causal inference is a major concern in published longitudinal studies evaluating impaired neurological function in children prenatally exposed to background levels of methyl mercury (MeHg). These studies evaluate a large number of developmental tests. We propose an alternative analysis strategy that reduces the number of comparisons tested in these studies.

METHODS

Using data from the 9-year follow-up of 643 children in the Seychelles child development study, we grouped 18 individual endpoints into one overall ordinal outcome variable as well as by developmental domains. Subsequently, ordinal logistic regression analyses were performed.

RESULTS

We did not find an association between prenatal MeHg exposure and developmental outcomes at 9 years of age.

CONCLUSION

Our proposed framework is more likely to result in a balanced interpretation of a posteriori associations. In addition, this new strategy should facilitate the use of complex epidemiological data in quantitative risk assessment.