Five multivariate Duchenne muscular dystrophy progression models bridging six-minute walk distance and MRI relaxometry of leg muscles

The study aimed to provide quantitative information on the utilization of MRI transverse relaxation time constant (MRI-T2) of leg muscles in DMD clinical trials by developing multivariate disease progression models of Duchenne muscular dystrophy (DMD) using 6-min walk distance (6MWD) and MRI-T2. Clinical data were collected from the prospective and longitudinal ImagingNMD study. Disease progression models were developed by a nonlinear mixed-effect modeling approach. Univariate models of 6MWD and MRI-T2 of five muscles were developed separately. Age at assessment was the time metric. Multivariate models were developed by estimating the correlation of 6MWD and MRI-T2 model variables. Full model estimation approach for covariate analysis and five-fold cross validation were conducted. Simulations were performed to compare the models and predict the covariate effects on the trajectories of 6MWD and MRI-T2. Sigmoid Imax and Emax models best captured the profiles of 6MWD and MRI-T2 over age. Steroid use, baseline 6MWD, and baseline MRI-T2 were significant covariates. The median age at which 6MWD is half of its maximum decrease in the five models was similar, while the median age at which MRI-T2 is half of its maximum increase varied depending on the type of muscle. The models connecting 6MWD and MRI-T2 successfully quantified how individual characteristics alter disease trajectories. The models demonstrate a plausible correlation between 6MWD and MRI-T2, supporting the use of MRI-T2. The developed models will guide drug developers in using the MRI-T2 to most efficient use in DMD clinical trials.

Short- and long-term safety of discontinuing chronic opioid therapy among older adults with Alzheimer's disease and related dementia

BACKGROUND

Limited evidence exists on the short- and long-term safety of discontinuing versus continuing chronic opioid therapy (COT) among patients with Alzheimer's disease and related dementias (ADRD).

METHODS

This cohort study was conducted among 162,677 older residents with ADRD and receipt of COT using a 100% Medicare nursing home sample. Discontinuation of COT was defined as no opioid refills for ≥90 days. Primary outcomes were rates of pain-related hospitalisation, pain-related emergency department visit, injury, opioid use disorder (OUD) and opioid overdose (OD) measured by diagnosis codes at quarterly intervals during 1- and 2-year follow-ups. Poisson regression models were fit using generalised estimating equations with inverse probability of treatment weights to model quarterly outcome rates between residents who discontinued versus continued COT.

RESULTS

The study sample consisted of 218,040 resident episodes with COT; of these episodes, 180,916 residents (83%) continued COT, whereas 37,124 residents (17%) subsequently discontinued COT. Discontinuing (vs. continuing) COT was associated with higher rates of all outcomes in the first quarter, but these associations attenuated over time. The adjusted rates of injury, OUD and OD were 0, 69 and 60% lower at the 1-year follow-up and 11, 81 and 79% lower at the 2-year follow-up, respectively, for residents who discontinued versus continued COT, with no difference in the adjusted rates of pain-related hospitalisations or emergency department visits.

CONCLUSIONS

The rates of adverse outcomes were higher in the first quarter but lower or non-differential at 1-year and 2-year follow-ups between COT discontinuers versus continuers among older residents with ADRD.

A Bayesian nonparametric approach for multiple mediators with applications in mental health studies

Mediation analysis with contemporaneously observed multiple mediators is a significant area of causal inference. Recent approaches for multiple mediators are often based on parametric models and thus may suffer from model misspecification. Also, much of the existing literature either only allow estimation of the joint mediation effect or estimate the joint mediation effect just as the sum of individual mediator effects, ignoring the interaction among the mediators. In this article, we propose a novel Bayesian nonparametric method that overcomes the two aforementioned drawbacks. We model the joint distribution of the observed data (outcome, mediators, treatment, and confounders) flexibly using an enriched Dirichlet process mixture with three levels. We use standardization (g-computation) to compute all possible mediation effects, including pairwise and all other possible interaction among the mediators. We thoroughly explore our method via simulations and apply our method to a mental health data from Wisconsin Longitudinal Study, where we estimate how the effect of births from unintended pregnancies on later life mental depression (CES-D) among the mothers is mediated through lack of self-acceptance and autonomy, employment instability, lack of social participation, and increased family stress. Our method identified significant individual mediators, along with some significant pairwise effects.

Pain intensity, physical function, and depressive symptoms associated with discontinuing long-term opioid therapy in older adults with Alzheimer's disease and related dementias

INTRODUCTION

Limited evidence exists on the associations of discontinuing versus continuing long-term opioid therapy (LTOT) with pain intensity, physical function, and depression among patients with Alzheimer's disease and related dementias (ADRD).

METHODS

A cohort study among 138,059 older residents with mild-to-moderate ADRD and receipt of LTOT was conducted using a 100% Medicare nursing home sample. Discontinuation of LTOT was defined as no opioid refills for ≥ 60 days. Outcomes were worsening pain, physical function, and depression from baseline to quarterly assessments during 1- and 2-year follow-ups.

RESULTS

The adjusted odds of worsening pain and depressive symptoms were 29% and 5% lower at the 1-year follow-up and 35% and 9% lower at the 2-year follow-up for residents who discontinued versus continued LTOT, with no difference in physical function.

DISCUSSION

Discontinuing LTOT was associated with lower short- and long-term worsening pain and depressive symptoms than continuing LTOT among older residents with ADRD.

HIGHLIGHTS

Discontinuing long-term opioid therapy (LTOT) was associated with lower short- and long-term worsening pain. Discontinuing LTOT was related to lower short- and long-term worsening depression. Discontinuing LTOT was not associated with short- and long-term physical function.

Clinical importance of changes in magnetic resonance biomarkers for Duchenne muscular dystrophy

OBJECTIVE

Magnetic resonance (MR) measures of muscle quality are highly sensitive to disease progression and predictive of meaningful functional milestones in Duchenne muscular dystrophy (DMD). This investigation aimed to establish the reproducibility, responsiveness to disease progression, and minimum clinically important difference (MCID) for multiple MR biomarkers at different disease stages in DMD using a large natural history dataset.

METHODS

Longitudinal MR imaging and spectroscopy outcomes and ambulatory function were measured in 180 individuals with DMD at three sites, including repeated measurements on two separate days (within 1 week) in 111 participants. These data were used to calculate day-to-day reproducibility, responsiveness (standardized response mean, SRM), minimum detectable change, and MCID. A survey of experts was also performed.

RESULTS

MR spectroscopy fat fraction (FF), as well as MR imaging transverse relaxation time (MRI-T2 ), measures performed in multiple leg muscles, and had high reproducibility (Pearson's R > 0.95). Responsiveness to disease progression varied by disease stage across muscles. The average FF from upper and lower leg muscles was highly responsive (SRM > 0.9) in both ambulatory and nonambulatory individuals. MCID estimated from the distribution of scores, by anchoring to function, and via expert opinion was between 0.01 and 0.05 for FF and between 0.8 and 3.7 ms for MRI-T2 .

INTERPRETATION

MR measures of FF and MRI T2 are reliable and highly responsive to disease progression. The MCID for MR measures is less than or equal to the typical annualized change. These results confirm the suitability of these measures for use in DMD and potentially other muscular dystrophies.

Flexible evaluation of surrogacy in platform studies

Trial-level surrogates are useful tools for improving the speed and cost effectiveness of trials but surrogates that have not been properly evaluated can cause misleading results. The evaluation procedure is often contextual and depends on the type of trial setting. There have been many proposed methods for trial-level surrogate evaluation, but none, to our knowledge, for the specific setting of platform studies. As platform studies are becoming more popular, methods for surrogate evaluation using them are needed. These studies also offer a rich data resource for surrogate evaluation that would not normally be possible. However, they also offer a set of statistical issues including heterogeneity of the study population, treatments, implementation, and even potentially the quality of the surrogate. We propose the use of a hierarchical Bayesian semiparametric model for the evaluation of potential surrogates using nonparametric priors for the distribution of true effects based on Dirichlet process mixtures. The motivation for this approach is to flexibly model relationships between the treatment effect on the surrogate and the treatment effect on the outcome and also to identify potential clusters with differential surrogate value in a data-driven manner so that treatment effects on the surrogate can be used to reliably predict treatment effects on the clinical outcome. In simulations, we find that our proposed method is superior to a simple, but fairly standard, hierarchical Bayesian method. We demonstrate how our method can be used in a simulated illustrative example (based on the ProBio trial), in which we are able to identify clusters where the surrogate is, and is not useful. We plan to apply our method to the ProBio trial, once it is completed.

Joint modeling the frequency and duration of accelerometer-measured physical activity from a lifestyle intervention trial

Physical activity (PA) guidelines recommend that PA be accumulated in bouts of 10 minutes or more in duration. Recently, researchers have sought to better understand how participants in PA interventions increase their activity. Participants can increase their daily PA by increasing the number of PA bouts per day while keeping the duration of the bouts constant; they can keep the number of bouts constant but increase the duration of each bout; or participants can increase both the number of bouts and their duration. We propose a novel joint modeling framework for modeling PA bouts and their duration over time. Our joint model is comprised of two sub-models: a mixed-effects Poisson hurdle sub-model for the number of bouts per day and a mixed-effects location scale gamma regression sub-model to characterize the duration of the bouts and their variance. The model allows us to estimate how daily PA bouts and their duration vary together over the course of an intervention and by treatment condition and is specifically designed to capture the unique distributional features of bouted PA as measured by accelerometer: frequent measurements, zero-inflated bouts, and skewed bout durations. We apply our methods to the Make Better Choices study, a longitudinal lifestyle intervention trial to increase PA. We perform a simulation study to evaluate how well our model is able to estimate relationships between outcomes.

Dirichlet process mixture models for the analysis of repeated attempt designs

In longitudinal studies, it is not uncommon to make multiple attempts to collect a measurement after baseline. Recording whether these attempts are successful provides useful information for the purposes of assessing missing data assumptions. This is because measurements from subjects who provide the data after numerous failed attempts may differ from those who provide the measurement after fewer attempts. Previous models for these designs were parametric and/or did not allow sensitivity analysis. For the former, there are always concerns about model misspecification and for the latter, sensitivity analysis is essential when conducting inference in the presence of missing data. Here, we propose a new approach which minimizes issues with model misspecification by using Bayesian nonparametrics for the observed data distribution. We also introduce a novel approach for identification and sensitivity analysis. We re-analyze the repeated attempts data from a clinical trial involving patients with severe mental illness and conduct simulations to better understand the properties of our approach.

Multivariate modeling of magnetic resonance biomarkers and clinical outcome measures for Duchenne muscular dystrophy clinical trials

Although regulatory agencies encourage inclusion of imaging biomarkers in clinical trials for Duchenne muscular dystrophy (DMD), industry receives minimal guidance on how to use these biomarkers most beneficially in trials. This study aims to identify the optimal use of muscle fat fraction biomarkers in DMD clinical trials through a quantitative disease-drug-trial modeling and simulation approach. We simultaneously developed two multivariate models quantifying the longitudinal associations between 6-minute walk distance (6MWD) and fat fraction measures from vastus lateralis and soleus muscles. We leveraged the longitudinal individual-level data collected for 10 years through the ImagingDMD study. Age of the individuals at assessment was chosen as the time metric. After the longitudinal dynamic of each measure was modeled separately, the selected univariate models were combined using correlation parameters. Covariates, including baseline scores of the measures and steroid use, were assessed using the full model approach. The nonlinear mixed-effects modeling was performed in Monolix. The final models showed reasonable precision of the parameter estimates. Simulation-based diagnostics and fivefold cross-validation further showed the model's adequacy. The multivariate models will guide drug developers on using fat fraction assessment most efficiently using available data, including the widely used 6MWD. The models will provide valuable information about how individual characteristics alter disease trajectories. We will extend the multivariate models to incorporate trial design parameters and hypothetical drug effects to inform better clinical trial designs through simulation, which will facilitate the design of clinical trials that are both more inclusive and more conclusive using fat fraction biomarkers.

A Bayesian semiparametric approach for inference on the population partly conditional mean from longitudinal data with dropout

Studies of memory trajectories using longitudinal data often result in highly nonrepresentative samples due to selective study enrollment and attrition. An additional bias comes from practice effects that result in improved or maintained performance due to familiarity with test content or context. These challenges may bias study findings and severely distort the ability to generalize to the target population. In this study, we propose an approach for estimating the finite population mean of a longitudinal outcome conditioning on being alive at a specific time point. We develop a flexible Bayesian semiparametric predictive estimator for population inference when longitudinal auxiliary information is known for the target population. We evaluate the sensitivity of the results to untestable assumptions and further compare our approach to other methods used for population inference in a simulation study. The proposed approach is motivated by 15-year longitudinal data from the Betula longitudinal cohort study. We apply our approach to estimate lifespan trajectories in episodic memory, with the aim to generalize findings to a target population.

Prenatal weight and regional body composition trajectories and neonatal body composition: The NICHD Foetal Growth Studies

BACKGROUND

Gestational weight gain (GWG) and anthropometric trajectories may affect foetal programming and are potentially modifiable.

OBJECTIVES

To assess concomitant patterns of change in weight, circumferences and adiposity across gestation as an integrated prenatal exposure, and determine how they relate to neonatal body composition.

METHODS

Data are from a prospective cohort of singleton pregnancies (n = 2182) enrolled in United States perinatal centres, 2009-2013. Overall and by prepregnancy BMI group (overweight/obesity and healthy weight), joint latent trajectory models were fit with prenatal weight, mid-upper arm circumference (MUAC), triceps (TSF) and subscapular (SSF) skinfolds. Differences in neonatal body composition by trajectory class were assessed via weighted least squares.

RESULTS

Six trajectory patterns reflecting co-occurring changes in weight and MUAC, SSF and TSF across pregnancy were identified overall and by body mass index (BMI) group. Among people with a healthy weight BMI, some differences were observed for neonatal subcutaneous adipose tissue, and among individuals with overweight/obesity some differences in neonatal lean mass were found. Neonatal adiposity measures were higher among infants born to individuals with prepregnancy overweight/obesity.

CONCLUSIONS

Six integrated trajectory patterns of prenatal weight, subcutaneous adipose tissue and circumferences were observed that were minimally associated with neonatal body composition, suggesting a stronger influence of prepregnancy BMI.

Evaluating Genetic Modifiers of Duchenne Muscular Dystrophy Disease Progression Using Modeling and MRI

BACKGROUND AND OBJECTIVES

Duchenne muscular dystrophy (DMD) is a progressive muscle degenerative disorder with a well-characterized disease phenotype but considerable interindividual heterogeneity that is not well understood. The aim of this study was to evaluate the effects of dystrophin variations and genetic modifiers of DMD on rate and age of muscle replacement by fat.

METHODS

One hundred seventy-five corticosteroid treated participants from the ImagingDMD natural history study underwent repeated magnetic resonance spectroscopy (MRS) of the vastus lateralis (VL) and soleus (SOL) to determine muscle fat fraction (FF). MRS was performed annually in most instances; however, some individuals had additional visits at 3 or 6 monthss intervals. FF changes over time were modeled using nonlinear mixed effects to estimate disease trajectories based on the age that the VL or SOL reached half-maximum change in FF (mu) and the time required for FF change (sigma). Computed mu and sigma values were evaluated for dystrophin variations that have demonstrated the ability to lead to a mild phenotype as well as compared between different genetic polymorphism groups.

RESULTS

Participants with dystrophin gene deletions amenable to exon 8 skipping (n = 4) had minimal increases in SOL FF and had an increase in VL mu value by 4.4 years compared with a reference cohort (p = 0.039). Participants with nonsense variations within exons that may produce milder phenotypes (n = 11) also had minimal increases in SOL and VL FFs. No differences in estimated FF trajectories were seen for individuals amenable to exon 44 skipping (n = 10). Modeling of the SPP1, LTBP4, and thrombospondin-1 (THBS1) genetic modifiers did not result in significant differences in muscle FF trajectories between genotype groups (p > 0.05); however, trends were noted for the polymorphisms associated with long-range regulation of LTBP4 and THBS1 that deserve further follow-up.

DISCUSSION

The results of this study link the historically mild phenotypes seen in individuals amenable to exon 8 skipping and with certain nonsense variations with alterations in trajectories of lower extremity muscle replacement by fat.

Combined impact of Medicare's hospital pay for performance programs on quality and safety outcomes is mixed

BACKGROUND

Three major hospital pay for performance (P4P) programs were introduced by the Affordable Care Act and intended to improve the quality, safety and efficiency of care provided to Medicare beneficiaries. The financial risk to hospitals associated with Medicare's P4P programs is substantial. Evidence on the positive impact of these programs, however, has been mixed, and no study has assessed their combined impact. In this study, we examined the combined impact of Medicare's P4P programs on clinical areas and populations targeted by the programs, as well as those outside their focus.

METHODS

We used 2007-2016 Healthcare Cost and Utilization Project State Inpatient Databases for 14 states to identify hospital-level inpatient quality indicators (IQIs) and patient safety indicators (PSIs), by quarter and payer (Medicare vs. non-Medicare). IQIs and PSIs are standardized, evidence-based measures that can be used to track hospital quality of care and patient safety over time using hospital administrative data. The study period of 2007-2016 was selected to capture multiple years before and after introduction of program metrics. Interrupted time series was used to analyze the impact of the P4P programs on study outcomes targeted and not targeted by the programs. In sensitivity analyses, we examined the impact of these programs on care for non-Medicare patients.

RESULTS

Medicare P4P programs were not associated with consistent improvements in targeted or non-targeted quality and safety measures. Moreover, mortality rates across targeted and untargeted conditions were generally getting worse after the introduction of Medicare's P4P programs. Trends in PSIs were extremely mixed, with five outcomes trending in an expected (improving) direction, five trending in an unexpected (deteriorating) direction, and three with insignificant changes over time. Sensitivity analyses did not substantially alter these results.

CONCLUSIONS

Consistent with previous studies for individual programs, we detect minimal, if any, effect of Medicare's hospital P4P programs on quality and safety. Given the growing evidence of limited impact, the administrative cost of monitoring and enforcing penalties, and potential increase in mortality, CMS should consider redesigning their P4P programs before continuing to expand them.

Longitudinal changes in cardiac function in Duchenne muscular dystrophy population as measured by magnetic resonance imaging

BACKGROUND

The lack of dystrophin in cardiomyocytes in Duchenne muscular dystrophy (DMD) is associated with progressive decline in cardiac function eventually leading to death by 20-40 years of age. The aim of this prospective study was to determine rate of progressive decline in left ventricular (LV) function in Duchenne muscular dystrophy (DMD) over 5 years.

METHODS

Short axis cine and grid tagged images of the LV were acquired in individuals with DMD (n = 59; age = 5.3-18.0 years) yearly, and healthy controls at baseline (n = 16, age = 6.0-18.3 years) on a 3 T MRI scanner. Grid-tagged images were analyzed for composite circumferential strain (ℇcc%) and ℇcc% in six mid LV segments. Cine images were analyzed for left ventricular ejection fraction (LVEF), LV mass (LVM), end-diastolic volume (EDV), end-systolic volume (ESV), LV atrioventricular plane displacement (LVAPD), and circumferential uniformity ratio estimate (CURE). LVM, EDV, and ESV were normalized to body surface area for a normalized index of LVM (LVMI), EDV (EDVI) and ESV (ESVI).

RESULTS

At baseline, LV ℇcc% was significantly worse in DMD compared to controls and five of the six mid LV segments demonstrated abnormal strain in DMD. Longitudinal measurements revealed that ℇcc% consistently declined in individuals with DMD with the inferior segments being more affected. LVEF progressively declined between 3 to 5 years post baseline visit. In a multivariate analysis, the use of cardioprotective drugs trended towards positively impacting cardiac measures while loss of ambulation and baseline age were associated with negative impact. Eight out of 17 cardiac parameters reached a minimal clinically important difference with a threshold of 1/3 standard deviation.

CONCLUSION

The study shows a worsening of circumferential strain in dystrophic myocardium. The findings emphasize the significance of early and longitudinal assessment of cardiac function in DMD and identify early biomarkers of cardiac dysfunction to help design clinical trials to mitigate cardiac pathology. This study provides valuable non-invasive and non-contrast based natural history data of cardiac changes which can be used to design clinical trials or interpret the results of current trials aimed at mitigating the effects of decreased cardiac function in DMD.

Characterizing Expiratory Respiratory Muscle Degeneration in Duchenne Muscular Dystrophy Using MRI

BACKGROUND

Expiratory muscle weakness and impaired airway clearance are early signs of respiratory dysfunction in Duchenne muscular dystrophy (DMD), a degenerative muscle disorder in which muscle cells are damaged and replaced by fibrofatty tissue. Little is known about expiratory muscle pathology and its relationship to cough and airway clearance capacity; however, the level of muscle replacement by fat can be estimated using MRI and expressed as a fat fraction (FF).

RESEARCH QUESTION

How does abdominal expiratory muscle fatty infiltration change over time in DMD and relate to clinical expiratory function?

STUDY DESIGN AND METHODS

Individuals with DMD underwent longitudinal MRI of the abdomen to determine FF in the internal oblique, external oblique, and rectus abdominis expiratory muscles. FF data were used to estimate a model of expiratory muscle degeneration by using nonlinear mixed effects and a cumulative distribution function. FVC, maximal inspiratory and expiratory pressures, and peak cough flow were collected as clinical correlates to MRI.

RESULTS

Forty individuals with DMD (aged 6-18 years at baseline) participated in up to five visits over 36 months. Modeling estimated the internal oblique progresses most quickly and reached 50% replacement by fat at a mean patient age of 13.0 years (external oblique, 14.0 years; rectus abdominis, 16.2 years). Corticosteroid-untreated individuals (n = 4) reached 50% muscle replacement by fat 3 to 4 years prior to treated individuals. Individuals with mild clinical dystrophic phenotypes (n = 3) reached 50% muscle replacement by fat 4 to 5 years later than corticosteroid-treated individuals. Internal and external oblique FFs near 50% were associated with maximal expiratory pressures < 60 cm H₂O and peak cough flows < 270 L/min.

INTERPRETATION

These data improve understanding of the early phase of respiratory compromise in DMD, which typically presents as airway clearance dysfunction prior to the onset of hypoventilation, and links expiratory muscle fatty infiltration to pulmonary function measures.

Step Activity Monitoring in Boys with Duchenne Muscular Dystrophy and its Correlation with Magnetic Resonance Measures and Functional Performance

BACKGROUND

Muscles of boys with Duchenne muscular dystrophy (DMD) are progressively replaced by fatty fibrous tissues, and weakness leads to loss of ambulation (LoA). Step activity (SA) monitoring is a quantitative measure of real-world ambulatory function. The relationship between quality of muscle health and SA is unknown in DMD.

OBJECTIVE

To determine SA in steroid treated boys with DMD across various age groups, and to evaluate the association of SA with quality of muscle health and ambulatory function.

METHODS

Quality of muscle health was measured by magnetic resonance (MR) imaging transverse magnetization relaxation time constant (MRI-T2) and MR spectroscopy fat fraction (MRS-FF). SA was assessed via accelerometry, and functional abilities were assessed through clinical walking tests. Correlations between SA, MR, and functional measures were determined. A threshold value of SA was determined to predict the future LoA.

RESULTS

The greatest reduction in SA was observed in the 9- < 11years age group. SA correlated with all functional and MR measures.10m walk/run test had the highest correlation with SA. An increase in muscle MRI-T2 and MRS-FF was associated with a decline in SA. Two years prior to LoA, SA in boys with DMD was 32% lower than age matched boys with DMD who maintained ambulation for more than two-year period. SA monitoring can predict subsequent LoA in Duchenne, as a daily step count of 3200 at baseline was associated with LoA over the next two-years.

CONCLUSION

SA monitoring is a feasible and accessible tool to measure functional capacity in the real-world environment.

Adjusting for selection bias due to missing data in electronic health records-based research

While electronic health records data provide unique opportunities for research, numerous methodological issues must be considered. Among these, selection bias due to incomplete/missing data has received far less attention than other issues. Unfortunately, standard missing data approaches (e.g. inverse-probability weighting and multiple imputation) generally fail to acknowledge the complex interplay of heterogeneous decisions made by patients, providers, and health systems that govern whether specific data elements in the electronic health records are observed. This, in turn, renders the missing-at-random assumption difficult to believe in standard approaches. In the clinical literature, the collection of decisions that gives rise to the observed data is referred to as the data provenance. Building on a recently-proposed framework for modularizing the data provenance, we develop a general and scalable framework for estimation and inference with respect to regression models based on inverse-probability weighting that allows for a hierarchy of missingness mechanisms to better align with the complex nature of electronic health records data. We show that the proposed estimator is consistent and asymptotically Normal, derive the form of the asymptotic variance, and propose two consistent estimators. Simulations show that naïve application of standard methods may yield biased point estimates, that the proposed estimators have good small-sample properties, and that researchers may have to contend with a bias-variance trade-off as they consider how to handle missing data. The proposed methods are motivated by an on-going, electronic health records-based study of bariatric surgery.

Glucagon-Like Peptide 1 Receptor Agonists and Chronic Lower Respiratory Disease Exacerbations Among Patients With Type 2 Diabetes

OBJECTIVE

Emerging data from animal and human pilot studies suggest potential benefits of glucagon-like peptide 1 receptor agonists (GLP-1RA) on lung function. We aimed to assess the association of GLP-1RA and chronic lower respiratory disease (CLRD) exacerbation in a population with comorbid type 2 diabetes (T2D) and CLRD.

RESEARCH DESIGN AND METHODS

A new-user active-comparator analysis was conducted with use of a national claims database of beneficiaries with employer-sponsored health insurance spanning 2005-2017. We included adults with T2D and CLRD who initiated GLP-1RA or dipeptidyl peptidase 4 inhibitors (DPP-4I) as an add-on therapy to their antidiabetes regimen. The primary outcome was time to first hospital admission for CLRD. The secondary outcome was a count of any CLRD exacerbation associated with an inpatient or outpatient visit. We estimated incidence rates using inverse probability of treatment weighting for each study group and compared via risk ratios.

RESULTS

The study sample consisted of 4,150 GLP-1RA and 12,540 DPP-4I new users with comorbid T2D and CLRD. The adjusted incidence rate of first CLRD admission during follow-up was 10.7 and 20.3 per 1,000 person-years for GLP-1RA and DPP-4I users, respectively, resulting in an adjusted hazard ratio of 0.52 (95% CI 0.32-0.85). For the secondary outcome, the adjusted incidence rate ratio was 0.70 (95% CI 0.57-0.87).

CONCLUSIONS

GLP-1RA users had fewer CLRD exacerbations in comparison with DPP-4I users. Considering both plausible mechanistic pathways and this real-world evidence, potential beneficial effects of GLP-1RA may be considered in selection of an antidiabetes treatment regimen. Randomized clinical trials are warranted to confirm our findings.

Bayesian semi-parametric G-computation for causal inference in a cohort study with MNAR dropout and death

Causal inference with observational longitudinal data and time-varying exposures is often complicated by time-dependent confounding and attrition. The G-computation formula is one approach for estimating a causal effect in this setting. The parametric modeling approach typically used in practice relies on strong modeling assumptions for valid inference, and moreover depends on an assumption of missing at random, which is not appropriate when the missingness is missing not at random (MNAR) or due to death. In this work we develop a flexible Bayesian semi-parametric G-computation approach for assessing the causal effect on the subpopulation that would survive irrespective of exposure, in a setting with MNAR dropout. The approach is to specify models for the observed data using Bayesian additive regression trees, and then use assumptions with embedded sensitivity parameters to identify and estimate the causal effect. The proposed approach is motivated by a longitudinal cohort study on cognition, health, and aging, and we apply our approach to study the effect of becoming a widow on memory. We also compare our approach to several standard methods.

A Semiparametric Bayesian Approach to Dropout in Longitudinal Studies with Auxiliary Covariates

We develop a semiparametric Bayesian approach to missing outcome data in longitudinal studies in the presence of auxiliary covariates. We consider a joint model for the full data response, missingness and auxiliary covariates. We include auxiliary covariates to "move" the missingness "closer" to missing at random (MAR). In particular, we specify a semiparametric Bayesian model for the observed data via Gaussian process priors and Bayesian additive regression trees. These model specifications allow us to capture non-linear and non-additive effects, in contrast to existing parametric methods. We then separately specify the conditional distribution of the missing data response given the observed data response, missingness and auxiliary covariates (i.e. the extrapolation distribution) using identifying restrictions. We introduce meaningful sensitivity parameters that allow for a simple sensitivity analysis. Informative priors on those sensitivity parameters can be elicited from subject-matter experts. We use Monte Carlo integration to compute the full data estimands. Performance of our approach is assessed using simulated datasets. Our methodology is motivated by, and applied to, data from a clinical trial on treatments for schizophrenia.

Effect of Telehealth Extended Care for Maintenance of Weight Loss in Rural US Communities: A Randomized Clinical Trial

IMPORTANCE

Lifestyle interventions for obesity produce reductions in body weight that can decrease risk for diabetes and cardiovascular disease but are limited by suboptimal maintenance of lost weight and inadequate dissemination in low-resource communities.

OBJECTIVE

To evaluate the effectiveness of extended care programs for obesity management delivered remotely in rural communities through the US Cooperative Extension System.

DESIGN, SETTING, AND PARTICIPANTS

This randomized clinical trial was conducted from October 21, 2013, to December 21, 2018, in Cooperative Extension Service offices of 14 counties in Florida. A total of 851 individuals were screened for participation; 220 individuals did not meet eligibility criteria, and 103 individuals declined to participate. Of 528 individuals who initiated a 4-month lifestyle intervention, 445 qualified for randomization. Data were analyzed from August 22 to October 21, 2019.

INTERVENTIONS

Participants were randomly assigned to extended care delivered via individual or group telephone counseling or an education control program delivered via email. All participants received 18 modules with posttreatment recommendations for maintaining lost weight. In the telephone-based interventions, health coaches provided participants with 18 individual or group sessions focused on problem solving for obstacles to the maintenance of weight loss.

MAIN OUTCOMES AND MEASURES

The primary outcome was change in body weight from the conclusion of initial intervention (month 4) to final follow-up (month 22). An additional outcome was the proportion of participants achieving at least 10% body weight reduction at follow-up.

RESULTS

Among 445 participants (mean [SD] age, 55.4 [10.2] years; 368 [82.7%] women; 329 [73.9%] white), 149 participants (33.5%) were randomized to individual telephone counseling, 143 participants (32.1%) were randomized to group telephone counseling, and 153 participants (34.4%) were randomized to the email education control. Mean (SD) baseline weight was 99.9 (14.6) kg, and mean (SD) weight loss after the initial intervention was 8.3 (4.9) kg. Mean weight regains at follow-up were 2.3 (95% credible interval [CrI], 1.2-3.4) kg in the individual telephone counseling group, 2.8 (95% CrI, 1.4-4.2) kg for the group telephone counseling group, and 4.1 (95% CrI, 3.1-5.0) kg for the education control group, with a significantly smaller weight regain observed in the individual telephone counseling group vs control group (posterior probability >.99). A larger proportion of participants in the individual telephone counseling group achieved at least 10% weight reductions (31.5% [95% CrI, 24.1%-40.0%]) than in the control group (19.1% [95% CrI, 14.1%-24.9%]) (posterior probability >.99).

CONCLUSIONS AND RELEVANCE

This randomized clinical trial found that providing extended care for obesity management in rural communities via individual telephone counseling decreased weight regain and increased the proportion of participants who sustained clinically meaningful weight losses.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT02054624.

Upper and Lower Extremities in Duchenne Muscular Dystrophy Evaluated with Quantitative MRI and Proton MR Spectroscopy in a Multicenter Cohort

Background Upper extremity MRI and proton MR spectroscopy are increasingly considered to be outcome measures in Duchenne muscular dystrophy (DMD) clinical trials. Purpose To demonstrate the feasibility of acquiring upper extremity MRI and proton (¹H) MR spectroscopy measures of T2 and fat fraction in a large, multicenter cohort (ImagingDMD) of ambulatory and nonambulatory individuals with DMD; compare upper and lower extremity muscles by using MRI and ¹H MR spectroscopy; and correlate upper extremity MRI and ¹H MR spectroscopy measures to function. Materials and Methods In this prospective cross-sectional study, MRI and ¹H MR spectroscopy and functional assessment data were acquired from participants with DMD and unaffected control participants at three centers (from January 28, 2016, to April 24, 2018). T2 maps of the shoulder, upper arm, forearm, thigh, and calf were generated from a spin-echo sequence (repetition time msec/echo time msec, 3000/20-320). Fat fraction maps were generated from chemical shift-encoded imaging (eight echo times). Fat fraction and ¹H₂O T2 in the deltoid and biceps brachii were measured from single-voxel ¹H MR spectroscopy (9000/11-243). Groups were compared by using Mann-Whitney test, and relationships between MRI and ¹H MR spectroscopy and arm function were assessed by using Spearman correlation. Results This study evaluated 119 male participants with DMD (mean age, 12 years ± 3 [standard deviation]) and 38 unaffected male control participants (mean age, 12 years ± 3). Deltoid and biceps brachii muscles were different in participants with DMD versus control participants in all age groups by using quantitative T2 MRI (P < .001) and ¹H MR spectroscopy fat fraction (P < .05). The deltoid, biceps brachii, and triceps brachii were affected to the same extent (P > .05) as the soleus and medial gastrocnemius. Negative correlations were observed between arm function and MRI (T2: range among muscles, ρ = -0.53 to -0.73 [P < .01]; fat fraction, ρ = -0.49 to -0.70 [P < .01]) and ¹H MR spectroscopy fat fraction (ρ = -0.64 to -0.71; P < .01). Conclusion This multicenter study demonstrated early and progressive involvement of upper extremity muscles in Duchenne muscular dystrophy (DMD) and showed the feasibility of MRI and ¹H MR spectroscopy to track disease progression over a wide range of ages in participants with DMD. © RSNA, 2020 Online supplemental material is available for this article.

Modeling disease trajectory in Duchenne muscular dystrophy

OBJECTIVE

To quantify disease progression in individuals with Duchenne muscular dystrophy (DMD) using magnetic resonance biomarkers of leg muscles.

METHODS

MRI and magnetic resonance spectroscopy (MRS) biomarkers were acquired from 104 participants with DMD and 51 healthy controls using a prospective observational study design with patients with DMD followed up yearly for up to 6 years. Fat fractions (FFs) in vastus lateralis and soleus muscles were determined with ¹H MRS. MRI quantitative T₂ (qT₂) values were measured for 3 muscles of the upper leg and 5 muscles of the lower leg. Longitudinal changes in biomarkers were modeled with a cumulative distribution function using a nonlinear mixed-effects approach.

RESULTS

MRS FF and MRI qT₂ increased with DMD disease duration, with the progression time constants differing markedly between individuals and across muscles. The average age at half-maximal muscle involvement (μ) occurred 4.8 years earlier in vastus lateralis than soleus, and these measures were strongly associated with loss-of-ambulation age. Corticosteroid treatment was found to delay μ by 2.5 years on average across muscles, although there were marked differences between muscles with more slowly progressing muscles showing larger delay.

CONCLUSIONS

MRS FF and MRI qT₂ provide sensitive noninvasive measures of DMD progression. Modeling changes in these biomarkers across multiple muscles can be used to detect and monitor the therapeutic effects of corticosteroids on disease progression and to provide prognostic information on functional outcomes. This modeling approach provides a method to transform these MRI biomarkers into well-understood metrics, allowing concise summaries of DMD disease progression at individual and population levels.

CLINICALTRIALSGOV IDENTIFIER

NCT01484678.

MR biomarkers predict clinical function in Duchenne muscular dystrophy

OBJECTIVE

To investigate the potential of lower extremity magnetic resonance (MR) biomarkers to serve as endpoints in clinical trials of therapeutics for Duchenne muscular dystrophy (DMD) by characterizing the longitudinal progression of MR biomarkers over 48 months and assessing their relationship to changes in ambulatory clinical function.

METHODS

One hundred sixty participants with DMD were enrolled in this longitudinal, natural history study and underwent MR data acquisition of the lower extremity muscles to determine muscle fat fraction (FF) and MRI T₂ biomarkers of disease progression. In addition, 4 tests of ambulatory function were performed. Participants returned for follow-up data collection at 12, 24, 36, and 48 months.

RESULTS

Longitudinal analysis of the MR biomarkers revealed that vastus lateralis FF, vastus lateralis MRI T₂, and biceps femoris long head MRI T₂ biomarkers were the fastest progressing biomarkers over time in this primarily ambulatory cohort. Biomarker values tended to demonstrate a nonlinear, sigmoidal trajectory over time. The lower extremity biomarkers predicted functional performance 12 and 24 months later, and the magnitude of change in an MR biomarker over time was related to the magnitude of change in function. Vastus lateralis FF, soleus FF, vastus lateralis MRI T₂, and biceps femoris long head MRI T₂ were the strongest predictors of future loss of function, including loss of ambulation.

CONCLUSIONS

This study supports the strong relationship between lower extremity MR biomarkers and measures of clinical function, as well as the ability of MR biomarkers, particularly those from proximal muscles, to predict future ambulatory function and important clinical milestones.

CLINICALTRIALSGOV IDENTIFIER

NCT01484678.

A Bayesian parametric approach to handle missing longitudinal outcome data in trial-based health economic evaluations

Trial-based economic evaluations are typically performed on cross-sectional variables, derived from the responses for only the completers in the study, using methods that ignore the complexities of utility and cost data (e.g. skewness and spikes). We present an alternative and more efficient Bayesian parametric approach to handle missing longitudinal outcomes in economic evaluations, while accounting for the complexities of the data. We specify a flexible parametric model for the observed data and partially identify the distribution of the missing data with partial identifying restrictions and sensitivity parameters. We explore alternative non-ignorable missingness scenarios through different priors for the sensitivity parameters, calibrated on the observed data. Our approach is motivated by, and applied to, data from a trial assessing the cost-effectiveness of a new treatment for intellectual disability and challenging behaviour.

Cost-Effectiveness of Three Doses of a Behavioral Intervention to Prevent or Delay Type 2 Diabetes in Rural Areas

BACKGROUND

Rural Americans have higher prevalence of obesity and type 2 diabetes (T2D) than urban populations and more limited access to behavioral programs to promote healthy lifestyle habits. Descriptive evidence from the Rural Lifestyle Intervention Treatment Effectiveness trial delivered through local cooperative extension service offices in rural areas previously identified that behavioral modification with both nutrition education and coaching resulted in a lower program delivery cost per kilogram of weight loss maintained at 2-years compared with an education-only comparator intervention.

OBJECTIVE

This analysis extended earlier Rural Lifestyle Intervention Treatment Effectiveness trial research regarding weight loss outcomes to assess whether nutrition education with behavioral coaching delivered through cooperative extension service offices is cost-effective relative to nutrition education only in reducing T2D cases in rural areas.

DESIGN

A cost-utility analysis was conducted.

PARTICIPANTS/SETTING

Trial participants (n=317) from June 2008 through June 2014 were adults residing in rural Florida counties with a baseline body mass index between 30 and 45, but otherwise identified as healthy.

INTERVENTION

Trial participants were randomly assigned to low, moderate, or high doses of behavioral coaching with nutrition education (ie, 16, 32, or 48 sessions over 24 months) or a comparator intervention that included 16 sessions of nutrition education without coaching. Participant glycated hemoglobin level was measured at baseline and the end of the trial to assess T2D status.

MAIN OUTCOME MEASURES

T2D categories by treatment arm were used to estimate participants' expected annual health care expenditures and expected health-related utility measured as quality adjusted life years (ie, QALYs) over a 5-year time horizon. Discounted incremental costs and QALYs were used to calculate incremental cost-effectiveness ratios for each behavioral coaching intervention dose relative to the education-only comparator.

STATISTICAL ANALYSES PERFORMED

Using a third-party payer perspective, Markov transition matrices were used to model participant transitions between T2D states. Replications of the individual participant behavior were conducted using Monte Carlo simulation.

RESULTS

All three doses of the behavioral coaching intervention had lower expected total costs and higher estimated QALYs than the education-only comparator. The moderate dose behavioral coaching intervention was associated with higher estimated QALYs but was costlier than the low dose; the moderate dose was favored over the low dose with willingness to pay thresholds over $107,895/QALY. The low dose behavioral coaching intervention was otherwise favored.

CONCLUSIONS

Because most rural Americans live in counties with cooperative extension service offices, nutrition education with behavioral coaching programs similar to those delivered through this trial may be effective and efficient in preventing or delaying T2D-associated consequences of obesity for rural adults.

Impact of the Hospital-Acquired Conditions Initiative on Falls and Physical Restraints: A Longitudinal Study

BACKGROUND

The Centers for Medicare & Medicaid Services (CMS) implemented the Hospital-Acquired Conditions (HACs) Initiative in October 2008; the CMS no longer reimbursed hospitals for fall injury. The effects of this payment change on fall and fall injury rates are not well described, nor its effect on physical restraint use.

OBJECTIVE

The aim of this study was to examine the effects of the 2008 HACs Initiative on the rates of falls, injurious falls, and physical restraint use.

DESIGN/SETTING

This was a nine-year retrospective cohort study (July 2006-December 2015) involving 2,862 adult medical, medical-surgical, and surgical nursing units from 734 hospitals.

MEASUREMENTS

Annual rates of change in falls, injurious falls, and physical restraint use during the two years before the payment rule went into effect were compared with one-, four-, and seven-year rates of annual change after implementation, adjusting for unit- and facility-level covariates. Stratified analyses were conducted according to bed size and teaching status.

RESULTS

Compared with prior to the payment change, there was stable acceleration in the one-, four-, and seven-year annual rates of decline in falls as follows: -2.1% (-3.3%, -0.9%), -2.2% (-3.2%, -1.1%), and -2.2% (-3.4%, -1.0%) respectively. For injurious falls, there was an increasing acceleration in the annual declines, achieving statistical significance only at seven years post CMS change as follows: -3.2% (-5.5%, -1.0%). Physical restraint use prevalence decreased from 1.6% to 0.6%. Changes in the rates of falls, injurious falls, and restraint use varied according to hospital bed size and teaching status.

CONCLUSIONS AND RELEVANCE

Since the HACs Initiative, there was at best a modest decline in the rates of falls and injurious falls observed primarily in larger, major teaching hospitals. An increase in restraint use was not observed. Falls remain a difficult patient safety problem for hospitals, and further research is required to develop cost-effective, generalizable strategies for their prevention.

A sensitivity analysis approach for informative dropout using shared parameter models

Shared parameter models (SPMs) are a useful approach to addressing bias from informative dropout in longitudinal studies. In SPMs it is typically assumed that the longitudinal outcome process and the dropout time are independent, given random effects and observed covariates. However, this conditional independence assumption is unverifiable. Currently, sensitivity analysis strategies for this unverifiable assumption of SPMs are underdeveloped. In principle, parameters that can and cannot be identified by the observed data should be clearly separated in sensitivity analyses, and sensitivity parameters should not influence the model fit to the observed data. For SPMs this is difficult because it is not clear how to separate the observed data likelihood from the distribution of the missing data given the observed data (i.e., 'extrapolation distribution'). In this article, we propose a new approach for transparent sensitivity analyses for informative dropout that separates the observed data likelihood and the extrapolation distribution, using a typical SPM as a working model for the complete data generating mechanism. For this model, the default extrapolation distribution is a skew-normal distribution (i.e., it is available in a closed form). We propose anchoring the sensitivity analysis on the default extrapolation distribution under the specified SPM and calibrate the sensitivity parameters using the observed data for subjects who drop out. The proposed approach is used to address informative dropout in the HIV Epidemiology Research Study.

BAYESIAN METHODS FOR MULTIPLE MEDIATORS: RELATING PRINCIPAL STRATIFICATION AND CAUSAL MEDIATION IN THE ANALYSIS OF POWER PLANT EMISSION CONTROLS

Emission control technologies installed on power plants are a key feature of many air pollution regulations in the US. While such regulations are predicated on the presumed relationships between emissions, ambient air pollution, and human health, many of these relationships have never been empirically verified. The goal of this paper is to develop new statistical methods to quantify these relationships. We frame this problem as one of mediation analysis to evaluate the extent to which the effect of a particular control technology on ambient pollution is mediated through causal effects on power plant emissions. Since power plants emit various compounds that contribute to ambient pollution, we develop new methods for multiple intermediate variables that are measured contemporaneously, may interact with one another, and may exhibit joint mediating effects. Specifically, we propose new methods leveraging two related frameworks for causal inference in the presence of mediating variables: principal stratification and causal mediation analysis. We define principal effects based on multiple mediators, and also introduce a new decomposition of the total effect of an intervention on ambient pollution into the natural direct effect and natural indirect effects for all combinations of mediators. Both approaches are anchored to the same observed-data models, which we specify with Bayesian nonparametric techniques. We provide assumptions for estimating principal causal effects, then augment these with an additional assumption required for causal mediation analysis. The two analyses, interpreted in tandem, provide the first empirical investigation of the presumed causal pathways that motivate important air quality regulatory policies.

Measurement error correction and sensitivity analysis in longitudinal dietary intervention studies using an external validation study

In lifestyle intervention trials, where the goal is to change a participant's weight or modify their eating behavior, self-reported diet is a longitudinal outcome variable that is subject to measurement error. We propose a statistical framework for correcting for measurement error in longitudinal self-reported dietary data by combining intervention data with auxiliary data from an external biomarker validation study where both self-reported and recovery biomarkers of dietary intake are available. In this setting, dietary intake measured without error in the intervention trial is missing data and multiple imputation is used to fill in the missing measurements. Since most validation studies are cross-sectional, they do not contain information on whether the nature of the measurement error changes over time or differs between treatment and control groups. We use sensitivity analyses to address the influence of these unverifiable assumptions involving the measurement error process and how they affect inferences regarding the effect of treatment. We apply our methods to self-reported sodium intake from the PREMIER study, a multi-component lifestyle intervention trial.

A note on compatibility for inference with missing data in the presence of auxiliary covariates

Imputation and inference (or analysis) models that cannot be true simultaneously are frequently used in practice when missing outcomes are present. In these situations, the conclusions can be misleading depending on how "different" the implicit inference model, induced by the imputation model, is from the inference model actually used. We introduce model-based compatibility (MBC) and compare two MBC approaches to a non-MBC approach and explore the inferential validity of the latter in a simple case. In addition, we evaluate more complex cases through a series of simulation studies. Overall, we recommend caution when making inferences using a non-MBC analysis and point out when the inferential "cost" is the largest.

Handling Missing Data in Instrumental Variable Methods for Causal Inference

It is very common in instrumental variable studies for there to be missing instrument data. For example, in the Wisconsin Longitudinal Study one can use genotype data as a Mendelian randomization-style instrument, but this information is often missing when subjects do not contribute saliva samples, or when the genotyping platform output is ambiguous. Here we review missing-at-random assumptions one can use to identify instrumental variable causal effects, and discuss various approaches for estimation and inference. We consider likelihood-based methods, regression and weighting estimators, and doubly robust estimators. The likelihood-based methods yield the most precise inference, and are optimal under the model assumptions, while the doubly robust estimators can attain the nonparametric efficiency bound while allowing flexible nonparametric estimation of nuisance functions (e.g., instrument propensity scores). The regression and weighting estimators can sometimes be easiest to describe and implement. Our main contribution is an extensive review of this wide array of estimators under varied missing-at-random assumptions, along with discussion of asymptotic properties and inferential tools. We also implement many of the estimators in an analysis of the Wisconsin Longitudinal Study, to study effects of impaired cognitive functioning on depression.

Classification using ensemble learning under weighted misclassification loss

Binary classification rules based on covariates typically depend on simple loss functions such as zero-one misclassification. Some cases may require more complex loss functions. For example, individual-level monitoring of HIV-infected individuals on antiretroviral therapy requires periodic assessment of treatment failure, defined as having a viral load (VL) value above a certain threshold. In some resource limited settings, VL tests may be limited by cost or technology, and diagnoses are based on other clinical markers. Depending on scenario, higher premium may be placed on avoiding false-positives, which brings greater cost and reduced treatment options. Here, the optimal rule is determined by minimizing a weighted misclassification loss/risk. We propose a method for finding and cross-validating optimal binary classification rules under weighted misclassification loss. We focus on rules comprising a prediction score and an associated threshold, where the score is derived using an ensemble learner. Simulations and examples show that our method, which derives the score and threshold jointly, more accurately estimates overall risk and has better operating characteristics compared with methods that derive the score first and the cutoff conditionally on the score especially for finite samples.

Effect of dose of behavioral weight loss treatment on glycemic control in adults with prediabetes

OBJECTIVE

This study examined the effects of three doses of behavioral weight loss treatment, compared with a nutrition education control group, on changes in glycemic control in individuals with obesity and prediabetes.

RESEARCH DESIGN AND METHODS

The study included 287 adults (77% female, 81% White; mean (SD) age=54.1 (10.5) years, body mass index=36.3 (3.9) kg/m², and hemoglobin A1c (HbA1c)=5.9 (0.2%)). Participants were randomized to one of three behavioral treatment doses (high=24 sessions, moderate=16 sessions, or low=8 sessions) or to an education group (control=8 sessions). Changes in HbA1c, fasting glucose, and body weight were assessed from baseline to 6 months.

RESULTS

Mean (99.2% credible interval (CI)) reductions in HbA1c were 0.11% (0.07% to 0.16%), 0.08% (0.03% to 0.13%), 0.03% (-0.01% to 0.07%), and 0.02% (-0.02% to 0.07%), for the high, moderate, low, and control conditions, respectively. Mean (CI) reductions in fasting blood glucose were 0.26 mmol/L (0.14 to 0.39), 0.09 mmol/L (0 to 0.19), 0.01 mmol/L (-0.07 to 0.09), and 0.04 mmol/L (-0.03 to 0.12) for the high, moderate, low, and control conditions, respectively. The high-dose treatment produced significantly greater reductions in HbA1c and fasting blood glucose than the low-dose and control conditions (posterior probabilities (pp)<0.001); no other significant between-group differences were observed. Mean (CI) reductions in body weight were 10.91 kg (9.30 to 12.64), 10.08 kg (8.38 to 11.72), 6.35 kg (5.19 to 7.69), and 3.82 kg (3.04 to 4.54) for the high, moderate, low, and control conditions, respectively. All between-group differences in 6-month weight change were significant (pps<0.001) except for the high-dose versus moderate-dose comparison.

CONCLUSION

For adults with obesity and prediabetes a high dose of behavioral treatment involving 24 sessions over 6 months may be needed to optimize improvements in glycemic control.

TRIAL REGISTRATION NUMBER

NCT00912652.

Bayesian nonparametric generative models for causal inference with missing at random covariates

We propose a general Bayesian nonparametric (BNP) approach to causal inference in the point treatment setting. The joint distribution of the observed data (outcome, treatment, and confounders) is modeled using an enriched Dirichlet process. The combination of the observed data model and causal assumptions allows us to identify any type of causal effect-differences, ratios, or quantile effects, either marginally or for subpopulations of interest. The proposed BNP model is well-suited for causal inference problems, as it does not require parametric assumptions about the distribution of confounders and naturally leads to a computationally efficient Gibbs sampling algorithm. By flexibly modeling the joint distribution, we are also able to impute (via data augmentation) values for missing covariates within the algorithm under an assumption of ignorable missingness, obviating the need to create separate imputed data sets. This approach for imputing the missing covariates has the additional advantage of guaranteeing congeniality between the imputation model and the analysis model, and because we use a BNP approach, parametric models are avoided for imputation. The performance of the method is assessed using simulation studies. The method is applied to data from a cohort study of human immunodeficiency virus/hepatitis C virus co-infected patients.

Optimizing and evaluating biomarker combinations as trial-level general surrogates

We extend the method proposed in a recent work by the Authors for trial-level general surrogate evaluation to allow combinations of biomarkers and provide a procedure for finding the "best" combination of biomarkers based on the absolute prediction error summary of surrogate quality. We use a nonparametric Bayesian model that allows us to select an optimal subset of biomarkers without having to consider a large number of explicit model specifications for that subset. This dramatically reduces the number of model comparisons needed. Given the model's flexibility, complex nonlinear relationships can be fit when enough data are available. We evaluate the operating characteristics of our proposed method in simulations designed to be similar to our motivating example. We use our method to compare and evaluate combinations of biomarkers as trial-level general surrogates for the pentavalent rotavirus vaccine RotaTeq™ (RV5) (Merck & Co, Inc, Kenilworth, New Jersey, USA), finding that the same single biomarker identified in our previously published analysis is likely the optimal subset.

Longitudinal timed function tests in Duchenne muscular dystrophy: ImagingDMD cohort natural history

INTRODUCTION

Tests of ambulatory function are common clinical trial endpoints in Duchenne muscular dystrophy (DMD). Using these tests, the ImagingDMD study has generated a large data set that can describe the contemporary natural history of DMD in 5-12.9-year-olds.

METHODS

Ninety-two corticosteroid-treated boys with DMD and 45 controls participated in this longitudinal study. Participants performed the 6-minute walk test (6MWT) and timed function tests (TFT: 10-m walk/run, climbing 4 stairs, supine to stand).

RESULTS

Boys with DMD had impaired functional performance even at 5-6.9 years old. Boys older than 7 had significant declines in function over 1 year for 10-m walk/run and 6MWT. Eighty percent of participants could perform all functional tests at 9 years old. TFTs appear to be slightly more responsive and predictive of disease progression than the 6MWT in 7-12.9 year olds.

DISCUSSION

This study provides insight into the contemporary natural history of key functional endpoints in DMD. Muscle Nerve 58: 631-638, 2018.

Bayesian Approaches for Missing Not at Random Outcome Data: The Role of Identifying Restrictions

Missing data is almost always present in real datasets, and introduces several statistical issues. One fundamental issue is that, in the absence of strong uncheckable assumptions, effects of interest are typically not nonparametrically identified. In this article, we review the generic approach of the use of identifying restrictions from a likelihood-based perspective, and provide points of contact for several recently proposed methods. An emphasis of this review is on restrictions for nonmonotone missingness, a subject that has been treated sparingly in the literature. We also present a general, fully-Bayesian, approach which is widely applicable and capable of handling a variety of identifying restrictions in a uniform manner.

Skeletal muscle magnetic resonance biomarkers correlate with function and sentinel events in Duchenne muscular dystrophy

OBJECTIVE

To provide evidence for quantitative magnetic resonance (qMR) biomarkers in Duchenne muscular dystrophy by investigating the relationship between qMR measures of lower extremity muscle pathology and functional endpoints in a large ambulatory cohort using a multicenter study design.

METHODS

MR spectroscopy and quantitative imaging were implemented to measure intramuscular fat fraction and the transverse magnetization relaxation time constant (T2) in lower extremity muscles of 136 participants with Duchenne muscular dystrophy. Measures were collected at 554 visits over 48 months at one of three imaging sites. Fat fraction was measured in the soleus and vastus lateralis using MR spectroscopy, while T2 was assessed using MRI in eight lower extremity muscles. Ambulatory function was measured using the 10m walk/run, climb four stairs, supine to stand, and six minute walk tests.

RESULTS

Significant correlations were found between all qMR and functional measures. Vastus lateralis qMR measures correlated most strongly to functional endpoints (|ρ| = 0.68-0.78), although measures in other rapidly progressing muscles including the biceps femoris (|ρ| = 0.63-0.73) and peroneals (|ρ| = 0.59-0.72) also showed strong correlations. Quantitative MR biomarkers were excellent indicators of loss of functional ability and correlated with qualitative measures of function. A VL FF of 0.40 was an approximate lower threshold of muscle pathology associated with loss of ambulation.

DISCUSSION

Lower extremity qMR biomarkers have a robust relationship to clinically meaningful measures of ambulatory function in Duchenne muscular dystrophy. These results provide strong supporting evidence for qMR biomarkers and set the stage for their potential use as surrogate outcomes in clinical trials.

A Bayesian nonparametric approach to causal inference on quantiles

We propose a Bayesian nonparametric approach (BNP) for causal inference on quantiles in the presence of many confounders. In particular, we define relevant causal quantities and specify BNP models to avoid bias from restrictive parametric assumptions. We first use Bayesian additive regression trees (BART) to model the propensity score and then construct the distribution of potential outcomes given the propensity score using a Dirichlet process mixture (DPM) of normals model. We thoroughly evaluate the operating characteristics of our approach and compare it to Bayesian and frequentist competitors. We use our approach to answer an important clinical question involving acute kidney injury using electronic health records.

ARMA Cholesky Factor Models for the Covariance Matrix of Linear Models

In longitudinal studies, serial dependence of repeated outcomes must be taken into account to make correct inferences on covariate effects. As such, care must be taken in modeling the covariance matrix. However, estimation of the covariance matrix is challenging because there are many parameters in the matrix and the estimated covariance matrix should be positive definite. To overcomes these limitations, two Cholesky decomposition approaches have been proposed: modified Cholesky decomposition for autoregressive (AR) structure and moving average Cholesky decomposition for moving average (MA) structure, respectively. However, the correlations of repeated outcomes are often not captured parsimoniously using either approach separately. In this paper, we propose a class of flexible, nonstationary, heteroscedastic models that exploits the structure allowed by combining the AR and MA modeling of the covariance matrix that we denote as ARMACD. We analyze a recent lung cancer study to illustrate the power of our proposed methods.

Bayesian methods for nonignorable dropout in joint models in smoking cessation studies

Inference on data with missingness can be challenging, particularly if the knowledge that a measurement was unobserved provides information about its distribution. Our work is motivated by the Commit to Quit II study, a smoking cessation trial that measured smoking status and weight change as weekly outcomes. It is expected that dropout in this study was informative and that patients with missed measurements are more likely to be smoking, even after conditioning on their observed smoking and weight history. We jointly model the categorical smoking status and continuous weight change outcomes by assuming normal latent variables for cessation and by extending the usual pattern mixture model to the bivariate case. The model includes a novel approach to sharing information across patterns through a Bayesian shrinkage framework to improve estimation stability for sparsely observed patterns. To accommodate the presumed informativeness of the missing data in a parsimonious manner, we model the unidentified components of the model under a non-future dependence assumption and specify departures from missing at random through sensitivity parameters, whose distributions are elicited from a subject-matter expert.

Associations among sugar sweetened beverage intake, visceral fat, and cortisol awakening response in minority youth

CONTEXT

Abdominal adiposity has long been associated with excess caloric intake possibly resulting from increased psychosocial stress and associated cortisol dysfunction. However, the relationship of sugar-sweetened beverage (SSB) intake specifically with cortisol variability and visceral adipose tissue (VAT) is unknown.

OBJECTIVE

To examine the relationships between SSB intake, VAT, and cortisol response in minority youth.

DESIGN

A cross-sectional analysis.

SETTING

The University of Southern California.

PARTICIPANTS

60 overweight/obese Non-Hispanic Black and Hispanic adolescents ages 14-18years.

MAIN OUTCOME MEASURES

VAT via Magnet Resonance Imaging (MRI), cortisol awakening response (CAR) via multiple salivary samples, and SSB intake via multiple 24-hour diet recalls. SSB intake was divided into the following: low SSB consumers (<1 servings per day), medium SSB consumers (≥1-<2 servings per day), high SSB consumers (≥2 servings per day). Analysis of covariance were run with VAT and CAR as dependent variables and SSB intake categories (independent variable) with the following a priori covariates: sex, Tanner stage, ethnicity, caloric intake, and body mass index.

RESULTS

The high SSB intake group exhibited a 7% higher VAT compared to the low SSB intake group (β=0.25, CI:(0.03, 0.33), p=0.02). CAR was associated with VAT (β=0.31, CI:(0.01,0.23), p=0.02). The high SSB intake group exhibited 22% higher CAR compared to the low SSB intake group (β=0.30, CI:(0.02,0.48), p=0.04).

CONCLUSION

This is the first study exploring the relationship between SSB, VAT, and CAR. SSB consumption appears to be independently associated greater abdominal adiposity and higher morning cortisol variability in overweight and obese minority youth. This study highlights potential targets for interventions specifically to reduce SSB intake in a minority youth population.

A framework for Bayesian nonparametric inference for causal effects of mediation

We propose a Bayesian non-parametric (BNP) framework for estimating causal effects of mediation, the natural direct, and indirect, effects. The strategy is to do this in two parts. Part 1 is a flexible model (using BNP) for the observed data distribution. Part 2 is a set of uncheckable assumptions with sensitivity parameters that in conjunction with Part 1 allows identification and estimation of the causal parameters and allows for uncertainty about these assumptions via priors on the sensitivity parameters. For Part 1, we specify a Dirichlet process mixture of multivariate normals as a prior on the joint distribution of the outcome, mediator, and covariates. This approach allows us to obtain a (simple) closed form of each marginal distribution. For Part 2, we consider two sets of assumptions: (a) the standard sequential ignorability (Imai et al., 2010) and (b) weakened set of the conditional independence type assumptions introduced in Daniels et al. (2012) and propose sensitivity analyses for both. We use this approach to assess mediation in a physical activity promotion trial.

Reevaluating Cumulative HIV-1 Viral Load as a Prognostic Predictor: Predicting Opportunistic Infection Incidence and Mortality in a Ugandan Cohort

Recent studies have evaluated cumulative human immunodeficiency virus type 1 (HIV-1) viral load (cVL) for predicting disease outcomes, with discrepant results. We reviewed the disparate methodological approaches taken and evaluated the prognostic utility of cVL in a resource-limited setting. Using data on the Infectious Diseases Institute (Makerere University, Kampala, Uganda) cohort, who initiated antiretroviral therapy in 2004-2005 and were followed up for 9 years, we calculated patients' time-updated cVL by summing the area under their viral load curves on either a linear scale (cVL1) or a logarithmic scale (cVL2). Using Cox proportional hazards models, we evaluated both metrics as predictors of incident opportunistic infections and mortality. Among 489 patients analyzed, neither cVL measure was a statistically significant predictor of opportunistic infection risk. In contrast, cVL2 (but not cVL1) was a statistically significant predictor of mortality, with each log10 increase corresponding to a 1.63-fold (95% confidence interval: 1.02, 2.60) elevation in mortality risk when cVL2 was accumulated from baseline. However, whether cVL is predictive or not hinges on difficult choices surrounding the cVL metric and statistical model employed. Previous studies may have suffered from confounding bias due to their focus on cVL1, which strongly correlates with other variables. Further methodological development is needed to illuminate whether the inconsistent predictive utility of cVL arises from causal relationships or from statistical artifacts.