Should regression calibration or multiple imputation be used when calibrating different devices in a longitudinal study?

In longitudinal studies, the devices used to measure exposures can change from visit to visit. Calibration studies, wherein a subset of participants is measured using both devices at follow-up, may be used to assess between-device differences (i.e., errors). Then, statistical methods are needed to adjust for between-device differences and the missing measurement data that often appear in calibration studies. Regression calibration and multiple imputation are two possible methods. We compared both methods in linear regression with a simulation study, considering various real-world scenarios for a longitudinal study of pulse wave velocity. Regression calibration and multiple imputation were both essentially unbiased, but correctly estimating the standard errors posed challenges. Multiple imputation with predicted mean matching produced close agreement with the empirical standard error. Fully stochastic multiple imputation underestimated the standard error by up to 50%, and regression calibration with bootstrapped standard errors performed slightly better than fully stochastic multiple imputation. Regression calibration was slightly more efficient than either multiple imputation method. The results suggest use of multiple imputation with predictive mean matching over fully stochastic imputation or regression calibration in longitudinal studies where a new device at follow-up might be error-prone compared to the device used at baseline.

Associations between avocado intake and measures of glucose and insulin homeostasis in Hispanic individuals with and without type 2 diabetes: Results from the Hispanic Community Health Study/Study of Latinos (HCHS/SOL).

BACKGROUND AND AIMS: To investigate associations between avocado intake and glycemia in adults with Hispanic/Latino ancestry. METHODS AND RESULTS: The associations of avocado intake with measures of insulin and glucose homeostasis were evaluated in a cross-sectional analysis of up to 14,591 Hispanic/Latino adults, using measures of: average glucose levels (hemoglobin A1c; HbA1c), fasting glucose and insulin, glucose and insulin levels after an oral glucose tolerance test (OGTT), and calculated measures of insulin resistance (HOMA-IR, and HOMA-%ß), and insulinogenic index. Associations were assessed using multivariable linear regression models, which controlled for sociodemographic factors and health behaviors, and which were stratified by dysglycemia status. In those with normoglycemia, avocado intake was associated with a higher insulinogenic index (ß = 0.17 ± 0.07, P = 0.02). In those with T2D (treated and untreated), avocado intake was associated with lower hemoglobin A1c (HbA1c; ß = -0.36 ± 0.21, P = 0.02), and lower fasting glucose (ß = -0.27 ± 0.12, P = 0.02). In the those with untreated T2D, avocado intake was additionally associated with HOMA-%ß (ß = 0.39 ± 0.19, P = 0.04), higher insulin values 2-h after an oral glucose load (ß = 0.62 ± 0.23, P = 0.01), and a higher insulinogenic index (ß = 0.42 ± 0.18, P = 0.02). No associations were observed in participants with prediabetes. CONCLUSIONS: We observed an association of avocado intake with better glucose/insulin homeostasis, especially in those with T2D.

Plasma metabolites associated with cognitive function across race/ethnicities affirming the importance of healthy nutrition.

INTRODUCTION: We studied the replication and generalization of previously identified metabolites potentially associated with global cognitive function in multiple race/ethnicities and assessed the contribution of diet to these associations. METHODS: We tested metabolite-cognitive function associations in U.S.A. Hispanic/Latino adults (n = 2222) from the Community Health Study/ Study of Latinos (HCHS/SOL) and in European (n = 1365) and African (n = 478) Americans from the Atherosclerosis Risk In Communities (ARIC) Study. We applied Mendelian Randomization (MR) analyses to assess causal associations between the metabolites and cognitive function and between Mediterranean diet and cognitive function. RESULTS: Six metabolites were consistently associated with lower global cognitive function across all studies. Of these, four were sugar-related (e.g., ribitol). MR analyses provided weak evidence for a potential causal effect of ribitol on cognitive function and bi-directional effects of cognitive performance on diet. DISCUSSION: Several diet-related metabolites were associated with global cognitive function across studies with different race/ethnicities. HIGHLIGHTS: Metabolites associated with cognitive function in Puerto Rican adults were recently identified. We demonstrate the generalizability of these associations across diverse race/ethnicities. Most identified metabolites are related to sugars. Mendelian Randomization (MR) provides weak evidence for a causal effect of ribitol on cognitive function. Beta-cryptoxanthin and other metabolites highlight the importance of a healthy diet.

Dynamic landmark prediction for mixture data.

In kin-cohort studies, clinicians want to provide their patients with the most current cumulative risk of death arising from a rare deleterious mutation. Estimating the cumulative risk is difficult when the genetic mutation status is unknown and only estimated probabilities of a patient having the mutation are available. We estimate the cumulative risk for this scenario using a novel nonparametric estimator that incorporates covariate information and dynamic landmark prediction. Our estimator has improved prediction accuracy over existing estimators that ignore covariate information. It is built within a dynamic landmark prediction framework whereby we can obtain personalized dynamic predictions over time. Compared to current standards, a simple transformation of our estimator provides more efficient estimates of marginal distribution functions in settings where patient-specific predictions are not the main goal. We show our estimator is unbiased and has more predictive accuracy compared to methods that ignore covariate information and landmarking. Applying our method to a Huntington disease study of mortality, we develop dynamic survival prediction curves incorporating gender and familial genetic information.

Estimating disease onset from change points of markers measured with error.

Huntington disease is an autosomal dominant, neurodegenerative disease without clearly identified biomarkers for when motor-onset occurs. Current standards to determine motor-onset rely on a clinician's subjective judgment that a patient's extrapyramidal signs are unequivocally associated with Huntington disease. This subjectivity can lead to error which could be overcome using an objective, data-driven metric that determines motor-onset. Recent studies of motor-sign decline-the longitudinal degeneration of motor-ability in patients-have revealed that motor-onset is closely related to an inflection point in its longitudinal trajectory. We propose a nonlinear location-shift marker model that captures this motor-sign decline and assesses how its inflection point is linked to other markers of Huntington disease progression. We propose two estimating procedures to estimate this model and its inflection point: one is a parametric method using nonlinear mixed effects model and the other one is a multi-stage nonparametric approach, which we developed. In an empirical study, the parametric approach was sensitive to correct specification of the mean structure of the longitudinal data. In contrast, our multi-stage nonparametric procedure consistently produced unbiased estimates regardless of the true mean structure. Applying our multi-stage nonparametric estimator to Neurobiological Predictors of Huntington Disease, a large observational study of Huntington disease, leads to earlier prediction of motor-onset compared to the clinician's subjective judgment.

Robust methods to correct for measurement error when evaluating a surrogate marker.

The identification of valid surrogate markers of disease or disease progression has the potential to decrease the length and costs of future studies. Most available methods that assess the value of a surrogate marker ignore the fact that surrogates are often measured with error. Failing to adjust for measurement error can erroneously identify a useful surrogate marker as not useful or vice versa. We investigate and propose robust methods to correct for the effect of measurement error when evaluating a surrogate marker using multiple estimators developed for parametric and nonparametric estimates of the proportion of treatment effect explained by the surrogate marker. In addition, we quantify the attenuation bias induced by measurement error and develop inference procedures to allow for variance and confidence interval estimation. Through a simulation study, we show that our proposed estimators correct for measurement error in the surrogate marker and that our inference procedures perform well in finite samples. We illustrate these methods by examining a potential surrogate marker that is measured with error, hemoglobin A1c, using data from the Diabetes Prevention Program clinical trial.

Correcting conditional mean imputation for censored covariates and improving usability.

Missing data are often overcome using imputation, which leverages the entire dataset to replace missing values with informed placeholders. This method can be modified for censored data by also incorporating partial information from censored values. One such modification proposed by Atem et al. (2017, 2019a, 2019b) is conditional mean imputation where censored covariates are replaced by their conditional means given other fully observed information. These methods are robust to additional parametric assumptions on the censored covariate and utilize all available data, which is appealing. However, in implementing these methods, we discovered that these three articles provide nonequivalent formulas and, in fact, none is the correct formula for the conditional mean. Herein, we derive the correct form of the conditional mean and discuss the bias incurred when using the incorrect formulas. Furthermore, we note that even the correct formula can perform poorly for log hazard ratios far from 0${\mathbf {0}}$ . We also provide user-friendly R software, the imputeCensoRd package, to enable future researchers to tackle censored covariates correctly.

svReg: Structural varying-coefficient regression to differentiate how regional brain atrophy affects motor impairment for Huntington disease severity groups.

For Huntington disease, identification of brain regions related to motor impairment can be useful for developing interventions to alleviate the motor symptom, the major symptom of the disease. However, the effects from the brain regions to motor impairment may vary for different groups of patients. Hence, our interest is not only to identify the brain regions but also to understand how their effects on motor impairment differ by patient groups. This can be cast as a model selection problem for a varying-coefficient regression. However, this is challenging when there is a pre-specified group structure among variables. We propose a novel variable selection method for a varying-coefficient regression with such structured variables and provide a publicly available R package svreg for implementation of our method. Our method is empirically shown to select relevant variables consistently. Also, our method screens irrelevant variables better than existing methods. Hence, our method leads to a model with higher sensitivity, lower false discovery rate and higher prediction accuracy than the existing methods. Finally, we found that the effects from the brain regions to motor impairment differ by disease severity of the patients. To the best of our knowledge, our study is the first to identify such interaction effects between the disease severity and brain regions, which indicates the need for customized intervention by disease severity.

Time-varying proportional odds model for mega-analysis of clustered event times.

Mega-analysis, or the meta-analysis of individual data, enables pooling and comparing multiple studies to enhance estimation and power. A challenge in mega-analysis is estimating the distribution for clustered, potentially censored event times where the dependency structure can introduce bias if ignored. We propose a new proportional odds model with unknown, time-varying coefficients, and random effects. The model directly captures event dependencies, handles censoring using pseudo-values, and permits a simple estimation by transforming the model into an easily estimable additive logistic mixed effect model. Our method consistently estimates the distribution for clustered event times even under covariate-dependent censoring. Applied to three observational studies of Huntington's disease, our method provides, for the first time in the literature, evidence of similar conclusions about motor and cognitive impairments in all studies despite different recruitment criteria.

A consistent estimator for logistic mixed effect models.

We propose a consistent and locally efficient estimator to estimate the model parameters for a logistic mixed effect model with random slopes. Our approach relaxes two typical assumptions: the random effects being normally distributed, and the covariates and random effects being independent of each other. Adhering to these assumptions is particularly difficult in health studies where in many cases we have limited resources to design experiments and gather data in long-term studies, while new findings from other fields might emerge, suggesting the violation of such assumptions. So it is crucial if we could have an estimator robust to such violations and then we could make better use of current data harvested using various valuable resources. Our method generalizes the framework presented in Garcia & Ma (2016) which also deals with a logistic mixed effect model but only considers a random intercept. A simulation study reveals that our proposed estimator remains consistent even when the independence and normality assumptions are violated. This contrasts from the traditional maximum likelihood estimator which is likely to be inconsistent when there is dependence between the covariates and random effects. Application of this work to a Huntington disease study reveals that disease diagnosis can be further improved using assessments of cognitive performance.

Statistical Approaches to Longitudinal Data Analysis in Neurodegenerative Diseases: Huntington's Disease as a Model.

Understanding the overall progression of neurodegenerative diseases is critical to the timing of therapeutic interventions and design of effective clinical trials. Disease progression can be assessed with longitudinal study designs in which outcomes are measured repeatedly over time and are assessed with respect to risk factors, either measured repeatedly or at baseline. Longitudinal data allows researchers to assess temporal disease aspects, but the analysis is complicated by complex correlation structures, irregularly spaced visits, missing data, and mixtures of time-varying and static covariate effects. We review modern statistical methods designed for these challenges. Among all methods, the mixed effect model most flexibly accommodates the challenges and is preferred by the FDA for observational and clinical studies. Examples from Huntington's disease studies are used for clarification, but the methods apply to neurodegenerative diseases in general, particularly as the identification of prodromal forms of neurodegenerative disease through sensitive biomarkers is increasing.

Simultaneous treatment of unspecified heteroskedastic model error distribution and mismeasured covariates for restricted moment models.

We develop consistent and efficient estimation of parameters in general regression models with mismeasured covariates. We assume the model error and covariate distributions are unspecified, and the measurement error distribution is a general parametric distribution with unknown variance-covariance. We construct root-n consistent, asymptotically normal and locally efficient estimators using the semiparametric efficient score. We do not estimate any unknown distribution or model error heteroskedasticity. Instead, we form the estimator under possibly incorrect working distribution models for the model error, error-prone covariate, or both. Empirical results demonstrate robustness to different incorrect working models in homoscedastic and heteroskedastic models with error-prone covariates.

Identification of important regressor groups, subgroups and individuals via regularization methods: application to gut microbiome data.

MOTIVATION: Gut microbiota can be classified at multiple taxonomy levels. Strategies to use changes in microbiota composition to effect health improvements require knowing at which taxonomy level interventions should be aimed. Identifying these important levels is difficult, however, because most statistical methods only consider when the microbiota are classified at one taxonomy level, not multiple. RESULTS: Using L1 and L2 regularizations, we developed a new variable selection method that identifies important features at multiple taxonomy levels. The regularization parameters are chosen by a new, data-adaptive, repeated cross-validation approach, which performed well. In simulation studies, our method outperformed competing methods: it more often selected significant variables, and had small false discovery rates and acceptable false-positive rates. Applying our method to gut microbiota data, we found which taxonomic levels were most altered by specific interventions or physiological status. AVAILABILITY: The new approach is implemented in an R package, which is freely available from the corresponding author. CONTACT: tpgarcia@srph.tamhsc.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Structured variable selection with q-values.

When some of the regressors can act on both the response and other explanatory variables, the already challenging problem of selecting variables when the number of covariates exceeds the sample size becomes more difficult. A motivating example is a metabolic study in mice that has diet groups and gut microbial percentages that may affect changes in multiple phenotypes related to body weight regulation. The data have more variables than observations and diet is known to act directly on the phenotypes as well as on some or potentially all of the microbial percentages. Interest lies in determining which gut microflora influence the phenotypes while accounting for the direct relationship between diet and the other variables A new methodology for variable selection in this context is presented that links the concept of q-values from multiple hypothesis testing to the recently developed weighted Lasso.

Understanding the implications of a complete case analysis for regression models with a right-censored covariate.

Despite its drawbacks, the complete case analysis is commonly used in regression models with incomplete covariates. Understanding when the complete case analysis will lead to consistent parameter estimation is vital before use. Our aim here is to demonstrate when a complete case analysis is consistent for randomly right-censored covariates and to discuss the implications of its use even when consistent. Across the censored covariate literature, different assumptions are made to ensure a complete case analysis produces a consistent estimator, which leads to confusion in practice. We make several contributions to dispel this confusion. First, we summarize the language surrounding the assumptions that lead to a consistent complete case estimator. Then, we show a unidirectional hierarchical relationship between these assumptions, which leads us to one sufficient assumption to consider before using a complete case analysis. Lastly, we conduct a simulation study to illustrate the performance of a complete case analysis with a right-censored covariate under different censoring mechanism assumptions, and we demonstrate its use with a Huntington disease data example.

Making Sense of Censored Covariates: Statistical Methods for Studies of Huntington's Disease.

The landscape of survival analysis is constantly being revolutionized to answer biomedical challenges, most recently the statistical challenge of censored covariates rather than outcomes. There are many promising strategies to tackle censored covariates, including weighting, imputation, maximum likelihood, and Bayesian methods. Still, this is a relatively fresh area of research, different from the areas of censored outcomes (i.e., survival analysis) or missing covariates. In this review, we discuss the unique statistical challenges encountered when handling censored covariates and provide an in-depth review of existing methods designed to address those challenges. We emphasize each method's relative strengths and weaknesses, providing recommendations to help investigators pinpoint the best approach to handling censored covariates in their data.

Association Between Hypertensive Disorders of Pregnancy and Interval Neurocognitive Decline: An Analysis of the Hispanic Community Health Study/Study of Latinos.

OBJECTIVE: To evaluate whether hypertensive disorders of pregnancy, including gestational hypertension, preeclampsia, and eclampsia, are associated with cognitive decline later in life among U.S. Hispanic/Latina individuals. METHODS: The HCHS/SOL (Hispanic Community Health Study/Study of Latinos) is a prospective population-based study of Hispanic/Latino individuals aged 18-74 years from four U.S. communities. This analysis included parous individuals aged 45 years or older who participated in the HCHS/SOL clinic study visit 1 (2008-2011) neurocognitive assessment and subsequently completed a repeat neurocognitive assessment as part of the Study of Latinos-Investigation of Neurocognitive Aging ancillary study visit 2 (2015-2018). Hypertensive disorders of pregnancy were assessed retrospectively by self-report of any gestational hypertension, preeclampsia, or eclampsia. Cognitive functioning was measured at both study visits with the Brief Spanish-English Verbal Learning Test, Digit Symbol Substitution, and Word Fluency. A regression-based approach was used to define cognitive decline at visit 2 as a function of cognition at visit 1 after adjustment for age, education, and follow-up time. Linear regression models were used to determine whether hypertensive disorders of pregnancy or their component diagnoses were associated with standardized cognitive decline after adjustment for sociodemographic characteristics, clinical and behavioral risk factors, and follow-up time. RESULTS: Among 3,554 individuals included in analysis, the mean age was 56.2 years, and 467 of individuals (13.4%) reported at least one hypertensive disorder of pregnancy. Individuals with hypertensive disorders of pregnancy compared with those without were more likely to have higher mean systolic blood pressure, fasting glucose, and body mass index. After an average of 7 years of follow-up, in fully adjusted models, gestational hypertension was associated with a 0.17-SD relative decline in Digit Symbol Substitution scores (95% CI, -0.31 to -0.04) but not other cognitive domains (Brief Spanish-English Verbal Learning Test or Word Fluency). Neither preeclampsia nor eclampsia was associated with neurocognitive differences. CONCLUSION: The presence of preeclampsia or eclampsia was not associated with interval neurocognitive decline. In this cohort of U.S. Hispanic/Latina individuals, gestational hypertension alone was associated with decreased processing speed and executive functioning later in life.

Genetic associations between sleep traits and cognitive ageing outcomes in the Hispanic Community Health Study/Study of Latinos.

BACKGROUND: Sleep phenotypes have been reported to be associated with cognitive ageing outcomes. However, there is limited research using genetic variants as proxies for sleep traits to study their associations. We estimated associations between Polygenic Risk Scores (PRSs) for sleep duration, insomnia, daytime sleepiness, and obstructive sleep apnoea (OSA) and measures of cogntive ageing in Hispanic/Latino adults. METHODS: We used summary statistics from published genome-wide association studies to construct PRSs representing the genetic basis of each sleep trait, then we studied the association of the PRSs of the sleep phenotypes with cognitive outcomes in the Hispanic Community Healthy Study/Study of Latinos. The primary model adjusted for age, sex, study centre, and measures of genetic ancestry. Associations are highlighted if their p-value <0.05. FINDINGS: Higher PRS for insomnia was associated with lower global cognitive function and higher risk of mild cognitive impairment (MCI) (OR = 1.20, 95% CI [1.06, 1.36]). Higher PRS for daytime sleepiness was also associated with increased MCI risk (OR = 1.14, 95% CI [1.02, 1.28]). Sleep duration PRS was associated with reduced MCI risk among short and normal sleepers, while among long sleepers it was associated with reduced global cognitive function and with increased MCI risk (OR = 1.40, 95% CI [1.10, 1.78]). Furthermore, adjustment of analyses for the measured sleep phenotypes and APOE-ε4 allele had minor effects on the PRS associations with the cognitive outcomes. INTERPRETATION: Genetic measures underlying insomnia, daytime sleepiness, and sleep duration are associated with MCI risk. Genetic and self-reported sleep duration interact in their effect on MCI. FUNDING: Described in Acknowledgments.

Association of neighborhood segregation with 6-year incidence of metabolic syndrome in the Hispanic community health study/study of Latinos.

PURPOSE: Examine the association between neighborhood segregation and 6-year incident metabolic syndrome (MetSyn) in the Hispanic Community Health Study/Study of Latinos. METHODS: Prospective cohort of adults residing in Miami, Chicago, the Bronx, and San Diego. The analytic sample included 6,710 participants who did not have MetSyn at baseline. The evenness and exposure dimensions of neighborhood segregation, based on the Gini and Isolation indices, respectively, were categorized into quintiles (Q). Racialized economic concentration was measured with the Index of Concentration at the Extremes (continuously and Q). RESULTS: Exposure, but not evenness, was associated with higher disease odds (Q1 (lower segregation) vs. Q4, OR = 1.53, 95% CI = 1.082.17; Q5, OR = 2.29, 95% CI = 1.493.52). Economic concentrationprivilege (continuous OR = 0.87, 95% CI = 0.770.98), racial concentrationracialized privilege (Q1 (greater concentration) vs. Q2 OR = 0.75, 95% CI = 0.541.04; Q3 OR = 0.68, 95% CI = 0.441.05; Q4 OR = 0.68, 95% CI = 0.451.01; Q5 OR = 0.64, 95% CI = 0.420.98)(continuous OR = 0.93, 95% CI = 0.821.04), and racialized economic concentrationprivilege (i.e., higher SES non-Hispanic White, continuous OR = 0.86, 95% CI = 0.760.98) were associated with lower disease odds. CONCLUSION: Hispanics/Latino adults residing in neighborhoods with high segregation had higher risk of incident MetSyn compared to those residing in neighborhoods with low segregation. Research is needed to identify the mechanisms that link segregation to poor metabolic health.

Efficiency improvement in a class of survival models through model-free covariate incorporation.

In randomized clinical trials, we are often concerned with comparing two-sample survival data. Although the log-rank test is usually suitable for this purpose, it may result in substantial power loss when the two groups have nonproportional hazards. In a more general class of survival models of Yang and Prentice (Biometrika 92:1-17, 2005), which includes the log-rank test as a special case, we improve model efficiency by incorporating auxiliary covariates that are correlated with the survival times. In a model-free form, we augment the estimating equation with auxiliary covariates, and establish the efficiency improvement using the semiparametric theories in Zhang et al. (Biometrics 64:707-715, 2008) and Lu and Tsiatis (Biometrics, 95:674-679, 2008). Under minimal assumptions, our approach produces an unbiased, asymptotically normal estimator with additional efficiency gain. Simulation studies and an application to a leukemia study show the satisfactory performance of the proposed method.