Random Mixed Models Research Articles

Speed of processing training and depression in assisted and independent living: A randomized controlled trial.

Late life depression is widely associated with lower quality of life and greater disability, making it an important target for prevention. Earlier randomized controlled trials [RCTs] demonstrated that speed of processing training [SOPT] led to reductions in depressive symptoms and clinical depression in community-dwelling adults. Our purpose was to evaluate depression outcomes related to SOPT among older adults who live in supported senior living settings. This two-arm, parallel RCT included 351 participants aged 55–102 years who resided in assisted and independent settings in 31 senior living communities. Participants were randomized within sites to computerized SOPT vs. computerized crossword puzzles with a targeted dose of 10 hours of playtime at baseline plus 4 hours of booster training at five and eleven months. Depression outcomes included the 9-item Patient Health Questionnaire [PHQ-9] scores, categorical levels, and dichotomous indicators. Random effects linear mixed effect models estimated SOPT effects in intention-to-treat complete case and multiple imputation analyses. Mean age of the sample was 81.0 years, 72.2% were women, and 41.0% resided in assisted living. At baseline 65.7% had no depression [PHQ-9 scores < 5] and 6.6% had clinically meaningful depression [PHQ-9 scores ≥ 10]. At 12 months we found significantly increased PHQ-9 scores [p = 0.006] and categorical levels [p = 0.003], and higher percentages of PHQ-2 scores > 3 [p = 0.016] and major depressive syndrome [p = 0.045] among the assisted living SOPT group. No significant change in depression was observed in the independent living SOPT or attention control groups. In summary, the SOPT known as Road Tour/Double Decision significantly increased, rather than decreased, the burden of depressive symptoms among participants residing in assisted living. Given these risks, this SOPT program should be avoided among older people in assisted living settings, and other SOPT interventions should be combined with systematic depression monitoring.

The effects of marijuana smoking on lung function in older people.

Previous studies have associated marijuana exposure with increased respiratory symptoms and chronic bronchitis among long-term cannabis smokers. The long-term effects of smoked marijuana on lung function remain unclear. We determined the association of marijuana smoking with the risk of spirometrically defined chronic obstructive pulmonary disease (COPD) (post-bronchodilator forced expiratory volume in 1 s (FEV1)/forced vital capacity ratio <0.7) in 5291 population-based individuals and the rate of decline in FEV1 in a subset of 1285 males and females, aged ≥40 years, who self-reported use (or non-use) of marijuana and tobacco cigarettes and performed spirometry before and after inhaled bronchodilator on multiple occasions. Analysis for the decline in FEV1 was performed using random mixed effects regression models adjusted for age, sex and body mass index. Heavy tobacco smoking and marijunana smoking was defined as >20 pack-years and >20 joint-years, respectively. ∼20% of participants had been or were current marijuana smokers with most having smoked tobacco cigarettes in addition (83%). Among heavy marijuana users, the risk of COPD was significantly increased (adjusted OR 2.45, 95% CI 1.55-3.88). Compared to never-smokers of marijuana and tobacco, heavy marijuana smokers and heavy tobacco smokers experienced a faster decline in FEV1 by 29.5 mL·year-1 (p=0.0007) and 21.1 mL·year-1 (p<0.0001), respectively. Those who smoked both substances experienced a decline of 32.31 mL·year-1 (p<0.0001). Heavy marijuana smoking increases the risk of COPD and accelerates FEV1 decline in concomitant tobacco smokers beyond that observed with tobacco alone.

Accounting for individual‐specific variation in habitat‐selection studies: Efficient estimation of mixed‐effects models using Bayesian or frequentist computation

Popular frameworks for studying habitat selection include resource-selection functions (RSFs) and step-selection functions (SSFs), estimated using logistic and conditional logistic regression, respectively. Both frameworks compare environmental covariates associated with locations animals visit with environmental covariates at a set of locations assumed available to the animals. Conceptually, slopes that vary by individual, that is, random coefficient models, could be used to accommodate inter-individual heterogeneity with either approach. While fitting such models for RSFs is possible with standard software for generalized linear mixed-effects models (GLMMs), straightforward and efficient one-step procedures for fitting SSFs with random coefficients are currently lacking. To close this gap, we take advantage of the fact that the conditional logistic regression model (i.e. the SSF) is likelihood-equivalent to a Poisson model with stratum-specific fixed intercepts. By interpreting the intercepts as a random effect with a large (fixed) variance, inference for random-slope models becomes feasible with standard Bayesian techniques, or with frequentist methods that allow one to fix the variance of a random effect. We compare this approach to other commonly applied alternatives, including models without random slopes and mixed conditional regression models fit using a two-step algorithm. Using data from mountain goats (Oreamnos americanus) and Eurasian otters (Lutra lutra), we illustrate that our models lead to valid and feasible inference. In addition, we conduct a simulation study to compare different estimation approaches for SSFs and to demonstrate the importance of including individual-specific slopes when estimating individual- and population-level habitat-selection parameters. By providing coded examples using integrated nested Laplace approximations (INLA) and Template Model Builder (TMB) for Bayesian and frequentist analysis via the R packages R-INLA and glmmTMB, we hope to make efficient estimation of RSFs and SSFs with random effects accessible to anyone in the field. SSFs with individual-specific coefficients are particularly attractive since they can provide insights into movement and habitat-selection processes at fine-spatial and temporal scales, but these models had previously been very challenging to fit.

Gait analysis with wearables predicts conversion to Parkinson disease

ObjectiveQuantification of gait with wearable technology is promising; recent cross‐sectional studies showed that gait characteristics are potential prodromal markers for Parkinson disease (PD). The aim of this longitudinal prospective observational study was to establish gait impairments and trajectories in the prodromal phase of PD, identifying which gait characteristics are potentially early diagnostic markers of PD.MethodsThe 696 healthy controls (mean age = 63 ± 7 years) recruited in the Tubingen Evaluation of Risk Factors for Early Detection of Neurodegeneration study were included. Assessments were performed longitudinally 4 times at 2‐year intervals, and people who converted to PD were identified. Participants were asked to walk at different speeds under single and dual tasking, with a wearable device placed on the lower back; 14 validated clinically relevant gait characteristics were quantified. Cox regression was used to examine whether gait at first visit could predict time to PD conversion after controlling for age and sex. Random effects linear mixed models (RELMs) were used to establish longitudinal trajectories of gait and model the latency between impaired gait and PD diagnosis.ResultsSixteen participants were diagnosed with PD on average 4.5 years after first visit (converters; PDC). Higher step time variability and asymmetry of all gait characteristics were associated with a shorter time to PD diagnosis. RELMs indicated that gait (lower pace) deviates from that of non‐PDC approximately 4 years prior to diagnosis.InterpretationTogether with other prodromal markers, quantitative gait characteristics can play an important role in identifying prodromal PD and progression within this phase. ANN NEUROL 2019;86:357–367

A hypothesis testing procedure for random changepoint mixed models.

In biomedical research, random changepoint mixed models are used to take into account an individual breakpoint in a biomarker trajectory. This may be observed in the cognitive decline measured by psychometric tests in the prediagnosis phase of Alzheimer's disease. The existence, intensity and duration of this accelerated decline can depend on individual characteristics. The main objective of our work is to propose inferential methods to assess the existence of this phase of accelerated decline, ie, the existence of a random changepoint. To do so, we use a mixed model with two linear phases and test the nullity of the parameter measuring the difference of slopes between the two phases. Because we face the issue of nuisance parameters being unidentifiable under the null hypothesis, the supremum of the classic score test statistic on these parameters is used. The asymptotic distribution of the supremum under the null is approached with a perturbation method based on the multiplier bootstrap. The performance of our testing procedure is assessed via simulations and the test is applied to the French cohort PAQUID of elderly subjects to study the shape of the prediagnosis decline according to educational level. The test is significant for both educational levels and the estimated trajectories confirmed that educational level is a good marker for cognitive reserve.

Effects Of Sit-to-stand Desk And Treadmill Workstations On Sedentary Behavior And Physical Activity

PURPOSE: To compare the efficacies of treadmill and sit-to-stand workstations in decreasing daily sedentary behavior (SED) during a 12-month, cluster-randomized trial with an intent to treat design in sedentary overweight office workers. METHODS: Sixty-six office workers (7 male, 59 female, age ± SD = 45.3 ± 12.3 y., BMI ± SD = 32.5 ± 5.7 kg/m2) were cluster randomized to one of 3 groups: (i) seated desk control (C) (N=21), (ii) sit-to-stand desk (D) (N=23), or (iii) treadmill desk (T) (N=22). Change in mean daily SED, standing and stepping time were measured using activPALTM accelerometers adhered to the dominant thigh at baseline (B), month-3 (M3), month-6 (M6) and month-12 (M12). Inclusion in analyses required ≥ 4 valid accelerometer wear days (i.e., ≥ 10 h. of wake wear time). Missing mean daily SED, standing and stepping hours were imputed using multiple-imputation. Between and within group differences in mean daily wake-time spent SED, standing, and stepping after M3, M6 and M12 were analyzed with random intercept mixed linear models accounting for repeated measures and clustering effects. Bonferroni corrections were used during pairwise post-hoc comparisons to correct for multiple hypotheses testing. RESULTS: Mean monitoring time (i.e., mean sensor wear days and daily time) did not significantly vary between or within groups. There were no significant within group changes in mean daily SED time. Mean daily standing time significantly increased in group D from B to M3 (Mean ± SD = 1.03 ± 2.24 h [7.72%], p = 0.03, Cohen’s D = 0.46), which was sustained through M12. Similarly, group T increased mean standing time from B to M3 (Mean ± SD = 1.23 ± 2.56 h [13.28 %], p = 0.025, Cohen’s D = 0.48), but these increases were negated at M12. Mean daily stepping time increased significantly in group D from B to M12 (Mean ± SD = 0.81 ± 1.61 h [2.65 %], p = 0.019, Cohen’s D = 0.50). No significant between group differences in SED, standing or stepping time were observed at M3, M6, or M12. CONCLUSION: Workstation-based workplace interventions may result in moderate short-term daily reductions in SED and increased physical activity among seated office workers. Sustaining these short-term behavioral improvements may not be achievable through passive environmental modifications alone, and may require additional active behavior change strategies.

Antiepileptic Drugs to Prevent Seizures After Spontaneous Intracerebral Hemorrhage.

Background and Purpose- We sought to evaluate the available literature to determine whether primary seizure prevention with antiepileptic drugs reduces the risk of poor outcomes and clinically relevant seizures among adult patients with spontaneous intracerebral hemorrhage. Methods- Meta-analysis of observational studies and randomized controlled trials evaluating the use of any antiepileptic drug for primary seizure prevention among adult (≥18 years) patients with spontaneous intracerebral hemorrhage. The primary end point was poor clinical outcome at the longest recorded follow-up, defined as either a high (>3) modified Rankin Scale score or all-cause mortality during follow-up if the modified Rankin Scale score was not recorded. Early and late seizures were secondary outcomes. A random mixed effects model was used to estimate the pooled odds ratio of outcomes and associated 95% CI. Results- We identified 7 studies with a total of 3241 patients for analysis of the primary outcome and 4 studies with a total of 1861 patients for analysis of the secondary outcomes. Overall, the use of antiepileptic drugs was not associated with a high Rankin Scale or all-cause mortality (odds ratio: 0.99; 95% CI, 0.66-1.49) or incident seizures (odds ratio: 0.89; 95% CI, 0.52-1.51) at the longest recorded follow-up time. Conclusions- The use of antiepileptic drugs as primary prevention among adult patients with spontaneous intracerebral hemorrhage is not associated with improved neurological function during long-term follow-up. Future studies should focus on the preventive use of distinct antiepileptic agents among patients at high risk of both seizures and poor outcomes.

DECELERATION PHASE MOMENTS: A POTENTIAL CAUSE FOR ELBOW INJURY IN COLLEGIATE LEVEL BASEBALL PITCHERS

Background: Over the past three decades there has been an increase in the incidence of elbow and shoulder pain experienced by baseball pitchers, which can limit or lead to an end of pitching activities. Although there are a number of theories that suggest poor pitching mechanics or throwing breaking pitches prior to skeletal maturity may be the cause, biomechanical investigations have yet to elucidate a single cause for this rise in injuries. It is also well established that the highest stresses and fastest angular velocities experienced by pitchers occurs during the acceleration phase of the pitch cycle, which has led to extensive biomechanical investigations of this portion of the pitching cycle. However, the deceleration phase of the pitch, although 150% longer than the acceleration phase, still requires an abrupt reversal of motion to allow pitchers to get into a fielding position after they have delivered the pitch. Therefore, the purpose of this study was to determine if the elbow joint was subjected to an additional varus stress during the deceleration phase of the pitch cycle. Methods: NCAA Division I and Division III baseball pitchers were recruited for this study and underwent a comprehensive biomechanical pitching evaluation. All pitchers were injury free at the time of data collection and reported no history of an upper extremity injury within the previous six months of the analysis date. Additionally, all pitchers had at least two years of pitching experience. All participants pitched from a 10” mound towards a target with a designated strike zone set 60’6” away. Kinematic data was collected using a 12-camera motion capture system, and kinetic data was calculated using standard inverse dynamic techniques. The typical pitching cycle, starting with lead foot contact and ending with maximum internal rotation of the glenohumeral joint (MIR), was expanded to end when the pedestal foot reached its maximum height; allowing for the analysis of deceleration phase moments at the elbow joint. The deceleration phase elbow varus (EV) moment was compared across multiple pitch types (i.e. fastball, curveball, slider, and change-up) using the type III effects from a random intercept mixed effects model. Additionally, the deceleration phase EV moment was compared to the peak EV moment occurring during the acceleration phase of the pitch cycle. Results: The results of this study are based on 87 baseball pitchers with a mean age of 19.9 ± 1.4 years. All participants pitched a fastball, 78 pitched a curveball, 31 pitched a slider, and 60 pitched a change-up. The results indicated that there was the presence of an elbow varus moment for all pitch types that occurred during the deceleration phase of the pitching cycle after MIR that was on average about half of the peak acceleration phase moment (Table 1). Overall 26% of pitchers pitching a fastball, 33% of pitchers throwing a curveball and change-up, and 55% of pitchers throwing a slider had deceleration EV moments greater than 50% of their peak acceleration phase EV moment. There was a statistically significant difference in the number of pitchers with a deceleration phase EV moment greater than half of the acceleration phase EV moment when pitching the slider when compared to the other pitch types (p=0.029). Conclusion/Significance: The majority of pitching biomechanics research focuses on the acceleration phase of the pitching cycle because the highest speeds and moments are achieved during this portion of the pitch. However, the pitcher’s need to rapidly decelerate during the pitch does expose them to an additional elbow varus moment. This additional moment could be a potential source of injury as it is a second stress exposure for the UCL. Additionally, given that the highest deceleration EV moments were noted in the slider this may potentially explain why pitchers and coaches believe that sliders are more harmful than other pitch types. [Table: see text]

Gaussian Markov Random Fields versus Linear Mixed Models for satellite-based PM2.5 assessment: Evidence from the Northeastern USA

Studying the effects of air-pollution on health is a key area in environmental epidemiology. An accurate estimation of air-pollution effects requires spatio-temporally resolved datasets of air-pollution, especially, Fine Particulate Matter (PM). Satellite-based technology has greatly enhanced the ability to provide PM assessments in locations where direct measurement is impossible.Indirect PM measurement is a statistical prediction problem. The spatio-temporal statistical literature offer various predictive models: Gaussian Random Fields (GRF) and Linear Mixed Models (LMM), in particular. GRF emphasize the spatio-temporal structure in the data, but are computationally demanding to fit. LMMs are computationally easier to fit, but require some tampering to deal with space and time.Recent advances in the spatio-temporal statistical literature propose to alleviate the computation burden of GRFs by approximating them with Gaussian Markov Random Fields (GMRFs). Since LMMs and GMRFs are both computationally feasible, the question arises: which is statistically better? We show that despite the great popularity of LMMs in environmental monitoring and pollution assessment, LMMs are statistically inferior to GMRF for measuring PM in the Northeastern USA.

How to analyse plant phenotypic plasticity in response to a changing climate.

Contents Summary 1235 I. Introduction 1235 II. The many shapes of phenotypic plasticity 1236 III. Random regression mixed model framework 1237 IV. Conclusions 1240 Acknowledgements 1240 References 1240 SUMMARY: Plant biology is experiencing a renewed interest in the mechanistic underpinnings and evolution of phenotypic plasticity that calls for a re-evaluation of how we analyse phenotypic responses to a rapidly changing climate. We suggest that dissecting plant plasticity in response to increasing temperature needs an approach that can represent plasticity over multiple environments, and considers both population-level responses and the variation between genotypes in their response. Here, we outline how a random regression mixed model framework can be applied to plastic traits that show linear or nonlinear responses to temperature. Random regressions provide a powerful and efficient means of characterising plasticity and its variation. Although they have been used widely in other fields, they have only recently been implemented in plant evolutionary ecology. We outline their structure and provide an example tutorial of their implementation.

Hot Topic: Global Trends in the Use of Opioid Analgesics

This study aimed to contribute to the debate on the availability of opioids for pain management by comparing regional and global trends for controlled strong opioids most commonly used for pain over the past 21 years. We conducted a series of analyses on the use of 12 opioids, with a focus on fentanyl, hydrocodone, hydromorphone, morphine, and oxycodone over the 21-year period of 1995–2015, at a regional and global level with country-level data, using random intercept mixed models to assess for change over time and interactions, and controlling for country economic development. Despite the sufficient supply of opioids to cover global demand, potent opioid pain medications are not available in many regions of the world. Prices of basic off-patent opioid formulations are significantly higher in poor countries as compared to lowest global prices, whilst countries may be under pressure to acquire expensive formulations. This results in lower availability of opioids for pain management in lower-income countries. The analysis shows that, when adjusted for country-level gross domestic product (GDP) per capita and Human Development Index (HDI) category, the most significant trends in opioid consumption for pain management are those of morphine and fentanyl. Whilst the trend for use of morphine has remained relatively stable, fentanyl use has risen sharply in the past 21 years, even in regions where the levels of overall opioid availability are inadequate or very inadequate. The trends vary according to different regions of the world.

Analysis of Hospitalization and Readmissions after CAR T Cell Therapy

Background: Chimeric antigen receptor (CAR) T-cell therapy is a rapidly emerging form of treatment for hematologic malignancies including lymphoma, multiple myeloma, and leukemia. Hospitalizations and readmissions after CAR T-cell therapy have not been systematically studied. A better understanding of hospital utilization patterns could inform design of clinical trials, advanced planning of hospitalizations, timing of discharge, and frequency and type of outpatient follow-up.Methods: We conducted a retrospective analysis of all patients admitted to the Massachusetts General Hospital for CAR T-cell therapy between 3/2016 and 3/2018. The primary outcome was hospital readmission within 30 days following discharge from CAR-T treatment visit. Secondary outcomes were ICU admission and inpatient mortality. Exploratory analyses were also conducted to determine whether patient age, length of initial hospitalization, laboratory measurements including ferritin and C-reactive protein (CRP), and tocilizumab exposure were associated with 30-day readmission. Summary statistics were reported by disease types, CAR-T products, and the primary and secondary outcomes using appropriate statistical functions. Comparison of laboratory values (ferritin and CRP) between the index admission (CAR T infusion) and first-readmission visits were analyzed using random intercept linear mixed effects models. Pairwise comparisons regarding mean difference of laboratory measurements during index admission and 30-day readmission/non-readmission were conducted and 95% confidence intervals reported. Bonferroni method was used for p-value adjustment (number of comparison groups = 4). This study was approved by the Partners Healthcare Institutional Review Board.Results: Forty-two patients were treated with CAR T-cells between 3/2016 and 3/2018. Thirty-six patients had non-Hodgkins lymphoma (NHL), 1 acute lymphoblastic leukemia, 1 chronic lymphocytic leukemia, and 4 multiple myeloma. Eight (19%) received the standard of care product, axicabtagene ciloleucel and all others were treated on a clinical trial with an investigational product. Median age at first treatment was 62.2 years old. Twenty-nine patients were male and 13 female. Fourteen patients (33.3%) experienced readmission within 30 days following discharge from CAR T infusion hospitalization. Twelve patients (29%) were readmitted within the first 14 days and 20 patients (47.6%) were readmitted to the hospital within 90 days of discharge (Figure 1). Patient age was not associated with 30-day readmission (P=0.642). Thirty-four (81%) patients received one CAR-T treatment and 8 (19%) received 2 treatments. Among those who received a second CAR-T treatment, 4 (50%) experienced 30-day readmission. Overall, 3 (7%) patients required transfer to the ICU (1 patient after both CAR T infusions) and two patients with NHL died during their first CAR T treatment admission. One patient died of progressive disease 44 days post-CAR T and the other died of disseminated candidemia 18 day post-CAR T. Our mixed effects model showed that median ferritin level was significantly elevated during the 30-day readmission visit as compared to initial CAR T infusion visit (1526.5 vs 1005.5 ug/L, difference in mean = 1088.5, 95% CI: 437.4 to 1739.6, adjusted P = 0.009). CAR T infusion visit ferritin levels did not differ among patients who were readmitted within 30 days and those who were not (1143.0 vs 1005.5 ug/L, P = 1.0). For readmissions which occurred after 30 days, we did not find a statistically significant increase in ferritin levels (P = 0.809). We did not find an association between initial length of hospitalization, CRP levels, and tocilizumab usage and early readmissions (all P-values > 0.05).Conclusions: CAR T-cell therapy is a promising evolving therapy for the treatment of relapsed, refractory hematologic malignancies. Further evaluation of pooled data may allow for early identification of patterns of deterioration which may limit premature hospital discharge, early readmission, and therapy-associated mortality. [Display omitted] DisclosuresNo relevant conflicts of interest to declare.

Bilateral Elective Nodal Radiation Increases Narcotic Requirement But Not Patient-Reported Pain During Head and Neck Cancer Radioation Therapy

Oral mucositis occurs in the majority of patients (pts) undergoing radiation therapy (RT) for head and neck cancer (HNC). The Oral Mucositis Daily Questionnaire (OMDQ) is a validated survey for HNC patient pain. This analysis was undertaken to report the course of patient-reported pain and associated narcotic use throughout the duration of RT for pts with HNC. Hypothesis: We hypothesized pts with bilateral elective nodal RT (BL-RT) treatment vs. unilateral elective nodal RT (UL-RT), with or without concurrent chemotherapy (CRT), would have significantly worse pain that coincided with increasing narcotic requirement. Pts with Stage I-IV HNC who were treated with either adjuvant or definitive RT at our institution from 2015 to 2017 were included for analysis. At baseline and weekly on-treatment visits, clinical data including narcotic use was recorded and OMDQ was administered. Pain was treated with either our institution’s standard of care (SOC) analgesic regimen using hydrocodone-acetaminophen +/- transdermal fentanyl as needed, or according to an outside provider (OP). Average daily narcotic use was converted to oral morphine milligram equivalents and reported as a cumulative requirement (dMME). Pts were dichotomized by either unilateral or BL-RT and +/- concurrent chemotherapy (RT vs CRT). To evaluate Mouth/Throat Soreness (MTS) OMDQ score response, the difference between baseline and week 3 (W3D) and at end-of-treatment (EoTD) was calculated for each pt. All other elements of the OMDQ were also analyzed. Two way ANOVA was performed and associations over time were quantified with a random coefficient mixed model. A total of 98 patients were eligible for analysis (42% SOC and 59% OP). The median age was 61.2y (IQR: 55.4-70.1y). Median dose to the primary was 70Gy (Range:50-70Gy). Overall, pts were treated: 44% adjuvantly, 56% definitively, 56% BL-RT, 39% UL-RT, 82% CRT, 18% RT alone. SOC vs. OP had no change in weekly MTS and no significant difference in dMME. There was no difference in overall pain scores, W3D or EoTD when comparing elective nodal RT; however, there was increased narcotic requirement for BL-RT (p<0.001). CRT vs. RT-alone had greater weekly MTS (p=0.03) and an increased dMME (p=0.047). There were no significant differences in all other OMDQ questions. We report on the use of narcotics and mucositis-related pain throughout the course of RT. Surprisingly, BL-RT did not increase MTS scores compared to UL-RT. Not surprisingly, CRT did increase weekly MTS scores. Both BL-RT and CRT increased narcotic requirements. Implications for research: This study suggests that narcotic use as measured by dMME, rather than OMDQ, may be a better metric for quantifying mucositis during radiation.

Latent trait shared-parameter mixed models for missing ecological momentary assessment data.

Latent trait shared-parameter mixed models for ecological momentary assessment (EMA) data containing missing values are developed in which data are collected in an intermittent manner. In such studies, data are often missing due to unanswered prompts. Using item response theory models, a latent trait is used to represent the missing prompts and modeled jointly with a mixed model for bivariate longitudinal outcomes. Both one- and two-parameter latent trait shared-parameter mixed models are presented. These new models offer a unique way to analyze missing EMA data with many response patterns. Here, the proposed models represent missingness via a latent trait that corresponds to the students' "ability" to respond to the prompting device. Data containing more than 10 300 observations from an EMA study involving high school students' positive and negative affects are presented. The latent trait representing missingness was a significant predictor of both positive affect and negative affect outcomes. The models are compared to a missing at random mixed model. A simulation study indicates that the proposed models can provide lower bias and increased efficiency compared to the standard missing at random approach commonly used with intermittent missing longitudinal data.

Data mining and machine learning approaches for the integration of genome-wide association and methylation data: methodology and main conclusions from GAW20

BackgroundMultiple layers of genetic and epigenetic variability are being simultaneously explored in an increasing number of health studies. We summarize here different approaches applied in the Data Mining and Machine Learning group at the GAW20 to integrate genome-wide genotype and methylation array data.ResultsWe provide a non-intimidating introduction to some frequently used methods to investigate high-dimensional molecular data and compare the different approaches tried by group members: random forest, deep learning, cluster analysis, mixed models, and gene-set enrichment analysis. Group contributions were quite heterogeneous regarding investigated data sets (real vs simulated), conducted data quality control and assessed phenotypes (eg, metabolic syndrome vs relative differences of log-transformed triglyceride concentrations before and after fenofibrate treatment). However, some common technical issues were detected, leading to practical recommendations.ConclusionsDifferent sources of correlation were identified by group members, including population stratification, family structure, batch effects, linkage disequilibrium and correlation of methylation values at neighboring cytosine-phosphate-guanine (CpG) sites, and the majority of applied approaches were able to take into account identified correlation structures. The ability to efficiently deal with high-dimensional omics data, and the model free nature of the approaches that did not require detailed model specifications were clearly recognized as the main strengths of applied methods. A limitation of random forest is its sensitivity to highly correlated variables. The parameter setup and the interpretation of results from deep learning methods, in particular deep neural networks, can be extremely challenging. Cluster analysis and mixed models may need some predimension reduction based on existing literature, data filtering, and supplementary statistical methods, and gene-set enrichment analysis requires biological insight.

A possible approach to improving the reproducibility of urinary concentrations of phthalate metabolites and phenols during pregnancy.

In epidemiologic studies, classifying episodic exposures to chemicals with short half-lives, such as phthalates and phenols, is challenging. We assessed whether accounting for sources of variability unrelated to exposure pathways would improve the reproducibility of urine concentrations of select phthalate metabolites and phenols. In 2011, a subset of pregnant women (n=19) enrolled in a prospective study provided first morning urines every 3-4 weeks between 16 and 36 weeks gestation. At the time of collection, we identified potential contributors to variations in urinary concentrations: weight gain, gestational age, time slept, time since awoke, time since last food/drink, and time since last void. We estimated intraclass correlation coefficients (ICCs) among repeat urine concentrations with and without adjustment for sources of variability using a random intercept linear mixed model. Concentrations of monoethyl phthalate, butyl- and propyl parabens were the most reproducible (ICCs: 0.68, 0.56, and 0.56 respectively). However, adjustment for potential sources of variability unrelated to exposure pathways did not materially improve reproducibility nor the ability of a single sample to predict exposure based on average biomarker concentrations across pregnancy. Future studies should carefully consider the exposure timeframe and the reliability of using biomarker concentrations from a single time point to represent exposures over pregnancy.

Pulmonary Circulation Transvascular Fluid Fluxes Do Not Change during General Anesthesia in Dogs.

General anesthesia (GA) can cause abnormal lung fluid redistribution. Pulmonary circulation transvascular fluid fluxes (JVA) are attributed to changes in hydrostatic forces and erythrocyte volume (EV) regulation. Despite the very low hydraulic conductance of pulmonary microvasculature it is possible that GA may affect hydrostatic forces through changes in pulmonary vascular resistance (PVR), and EV through alteration of erythrocyte transmembrane ion fluxes (ionJVA). Furosemide (Fur) was also used because of its potential to affect pulmonary hydrostatic forces and ionJVA. A hypothesis was tested that JVA, with or without furosemide treatment, will not change with time during GA. Twenty dogs that underwent castration/ovariectomy were randomly assigned to Fur (n = 10) (4 mg/kg IV) or placebo treated group (Con, n = 10). Baseline arterial (BL) and mixed venous blood were sampled during GA just before treatment with Fur or placebo and then at 15, 30 and 45 min post-treatment. Cardiac output (Q) and pulmonary artery pressure (PAP) were measured. JVA and ionJVA were calculated from changes in plasma protein, hemoglobin, hematocrit, plasma and whole blood ions, and Q. Variables were analyzed using random intercept mixed model (P < 0.05). Data are expressed as means ± SE. Furosemide caused a significant volume depletion as evident from changes in plasma protein and hematocrit (P < 0.001). However; Q, PAP, and JVA were not affected by time or Fur, whereas erythrocyte fluid flux was affected by Fur (P = 0.03). Furosemide also affected erythrocyte transmembrane K+ and Cl−, and transvascular Cl− metabolism (P ≤ 0.05). No other erythrocyte transmembrane or transvascular ion fluxes were affected by time of GA or Fur. Our hypothesis was verified as JVA was not affected by GA or ion metabolism changes due to Fur treatment. Furosemide and 45 min of GA did not cause significant hydrostatic changes based on Q and PAP. Inhibition of Na+/K+/2Cl− cotransport caused by Fur treatment, which can alter EV regulation and JVA, was offset by the Jacobs Stewart cycle. The results of this study indicate that the Jacobs Stewart cycle/erythrocyte Cl− metabolism can also act as a safety factor for the stability of lung fluid redistribution preserving optimal diffusion distance across the blood gas barrier.

A Monte Carlo study of REML and robust rank-based analyses for the random intercept mixed model

ABSTRACTRestricted maximum likelihood (REML) methods are traditionally used for analyzing mixed models. Based on a multivariate normal likelihood, these analyses are sensitive to outliers. Recently developed robust rank-based procedures offer a complete analysis of mixed model: estimation of fixed effects, standard errors, and estimation of variance components. The results of a large Monte Carlo study are presented, comparing these two analyses for many situations over multivariate normal and contaminated normal distributions. The rank-based analyses are much more powerful and efficient than the REML analyses over all non-normal situations, while losing little power for normal errors.

Hydrologic responses to restored wildfire regimes revealed by soil moisture-vegetation relationships

Abstract Many forested mountain watersheds worldwide evolved with frequent fire, which Twentieth Century fire suppression activities eliminated, resulting in unnaturally dense forests with high water demand. Restoration of pre-suppression forest composition and structure through a variety of management activities could improve forest resilience and water yields. This study explores the potential for “managed wildfire”, whereby naturally ignited fires are allowed to burn, to alter the water balance. Interest in this type of managed wildfire is increasing, yet its long-term effects on water balance are uncertain. We use soil moisture as a spatially-distributed hydrologic indicator to assess the influence of vegetation, fire history and landscape position on water availability in the Illilouette Creek Basin in Yosemite National Park. Over 6000 manual surface soil moisture measurements were made over a period of three years, and supplemented with continuous soil moisture measurements over the top 1m of soil in three sites. Random forest and linear mixed effects models showed a dominant effect of vegetation type and history of vegetation change on measured soil moisture. Contemporary and historical vegetation maps were used to upscale the soil moisture observations to the basin and infer soil moisture under fire-suppressed conditions. Little change in basin-averaged soil moisture was inferred due to managed wildfire, but the results indicated that large localized increases in soil moisture had occurred, which could have important impacts on local ecology or downstream flows.

Impact of non-uniform correlation structure on sample size and power in multiple-period cluster randomised trials.

Stepped wedge and cluster randomised crossover trials are examples of cluster randomised designs conducted over multiple time periods that are being used with increasing frequency in health research. Recent systematic reviews of both of these designs indicate that the within-cluster correlation is typically taken account of in the analysis of data using a random intercept mixed model, implying a constant correlation between any two individuals in the same cluster no matter how far apart in time they are measured: within-period and between-period intra-cluster correlations are assumed to be identical. Recently proposed extensions allow the within- and between-period intra-cluster correlations to differ, although these methods require that all between-period intra-cluster correlations are identical, which may not be appropriate in all situations. Motivated by a proposed intensive care cluster randomised trial, we propose an alternative correlation structure for repeated cross-sectional multiple-period cluster randomised trials in which the between-period intra-cluster correlation is allowed to decay depending on the distance between measurements. We present results for the variance of treatment effect estimators for varying amounts of decay, investigating the consequences of the variation in decay on sample size planning for stepped wedge, cluster crossover and multiple-period parallel-arm cluster randomised trials. We also investigate the impact of assuming constant between-period intra-cluster correlations instead of decaying between-period intra-cluster correlations. Our results indicate that in certain design configurations, including the one corresponding to the proposed trial, a correlation decay can have an important impact on variances of treatment effect estimators, and hence on sample size and power. An R Shiny app allows readers to interactively explore the impact of correlation decay.