Multilevel models for censored and latent responses

The impact of correlated exposures and missing data on multiple informant models used to identify critical exposure windows

There has been heightened interest in identifying critical windows of exposure for adverse health outcomes; that is, time points during which exposures have the greatest impact on a person's health. Multiple informant models implemented using generalized estimating equations (MIM GEEs) have been applied to address this research question because they enable statistical comparisons of differences in associations across exposure windows. As interest rises in using MIMs, the feasibility and appropriateness of their application under settings of correlated exposures and partially missing exposure measurements requires further examination. We evaluated the impact of correlation between exposure measurements and missing exposure data on the power and differences in association estimated by the MIM GEE and an inverse probability weighted extension to account for informatively missing exposures. We assessed these operating characteristics under a variety of correlation structures, sample sizes, and missing data mechanisms considering various exposure-outcome scenarios. We showed that applying MIM GEEs maintains higher power when there is a single critical window of exposure and exposure measures are not highly correlated, but may result in low power and bias under other settings. We applied these methods to a study of pregnant women living with HIV to explore differences in association between trimester-specific viral load and infant neurodevelopment.

Morning exercise mitigates the impact of prolonged sitting on cerebral blood flow in older adults

Preventing declines in cerebral blood flow is important for maintaining optimal brain health with aging. We compared the effects of a morning bout of moderate-intensity exercise, with and without subsequent light-intensity walking breaks from sitting, on cerebral blood velocity over 8 h in older adults. In a randomized crossover trial, overweight/obese older adults ( n = 12, 70 ± 7 yr; 30.4 ± 4.3 kg/m2), completed three acute conditions (6-day washout); SIT: prolonged sitting (8 h, control); EX+SIT: sitting (1 h), moderate-intensity walking (30 min), followed by uninterrupted sitting (6.5 h); and EX + BR: sitting (1 h), moderate-intensity walking (30 min), followed by sitting (6.5 h) interrupted with 3 min of light-intensity walking every 30 min. Bilateral middle cerebral artery velocities (MCAv) were determined using transcranial Doppler at 13 time points across the day. The temporal pattern and average MCAv over 8 h was determined. The pattern of MCAv over 8 h was a negative linear trend in SIT ( P < 0.001), but a positive quadratic trend in EX + SIT ( P < 0.001) and EX + BR ( P < 0.01). Afternoon time points in SIT were lower than baseline within condition ( P ≤ 0.001 for all). A morning dip in MCAv was observed in EX + SIT and EX + BR ( P < 0.05 relative to baseline), but afternoon time points were not significantly lower than baseline. The average MCAv over 8 h was higher in EX + SIT than SIT ( P = 0.007) or EX + BR ( P = 0.024). Uninterrupted sitting should be avoided, and moderate-intensity exercise should be encouraged for the daily maintenance of cerebral blood flow in older adults. The clinical implications of maintaining adequate cerebral blood flow include the delivery of vital oxygen and nutrients to the brain. NEW & NOTEWORTHY This is the first study to measure the combined effects of an exercise bout with breaks in sitting on cerebral blood velocity in older adults. Using frequent recordings over an 8-h period, we have performed a novel analysis of the pattern of cerebral blood velocity, adjusting for concurrent measures of mean arterial pressure and other potential confounders in a linear mixed effects regression.

Efficient Estimation for Semiparametric Structural Equation Models With Censored Data

Structural equation modeling is commonly used to capture complex structures of relationships among multiple variables, both latent and observed. We propose a general class of structural equation models with a semiparametric component for potentially censored survival times. We consider nonparametric maximum likelihood estimation and devise a combined Expectation-Maximization and Newton-Raphson algorithm for its implementation. We establish conditions for model identifiability and prove the consistency, asymptotic normality, and semiparametric efficiency of the estimators. Finally, we demonstrate the satisfactory performance of the proposed methods through simulation studies and provide an application to a motivating cancer study that contains a variety of genomic variables. Supplementary materials for this article are available online.

Intravitreal Bevacizumab Therapy for Choroidal Neovascularization Secondary to Age-Related Macular Degeneration: 6-Month Results of an Open-Label Uncontrolled Clinical Study

Purpose

To investigate the 6-month safety and clinical outcomes of intravitreal injections of bevacizumab administered to treat choroidal neovascularization secondary to age-related macular degeneration.

Methods

Twenty-seven patients underwent 1.25 mg intravitreal injections of bevacizumab at baseline. A similar intravitreal injection was administered to all eyes at 1 and 2 month follow-up visits. At baseline and at each follow-up visit (1, 2, 3, and 6 months), patients underwent best-corrected visual acuity (BCVA) measurement, fluorescein angiography, indocyanine green angiography, and optical coherence tomography. Laboratory testing, visual field analyses, and endothelial cell counts were performed at baseline and third and sixth months.

Results

At 3 months, the mean BCVA remained substantially stable at 20/100. Mean central retinal thickness (CRT) decreased from 373 to 279 μm (p<0.01). Mean lesion greatest linear dimension (GLD) decreased from 4087 to 3782 microns (p<0.01). At 6 months, mean BCVA slightly decreased from 20/100−1 to 20/125−3 (not significant, p=0.40). Mean CRT was still inferior to baseline (305 μm, p<0.01). Mean lesion GLD was 4186 μm, not different from baseline values (p=0.59), but superior to 3-month mean GLD (p<0.01). Significant visual field defects or endothelial cell losses were not detected at 3 and 6 months. Laboratory testing did not reveal any clinically significant deviations compared to baseline values.

Conclusions

Intravitreal therapy using bevacizumab over 6 months showed stabilization of visual acuity and choroidal neovascularization activity; the safety data were convincing. (Eur J Ophthalmol 2007; 17: 230–7)

Correcting for covariate measurement error in logistic regression using nonparametric maximum likelihood estimation

When covariates are measured with error, inference based on conventional generalized linear models can yield biased estimates of regression parameters. This problem can potentially be rectified by using generalized linear latent and mixed models (GLLAMM), including a measurement model for the relationship between observed and true covariates. However, the models are typically estimated under the assumption that both the true covariates and the measurement errors are normally distributed, although skewed covariate distributions are often observed in practice. In this article we relax the normality assumption for the true covariates by developing nonparametric maximum likelihood estimation (NPMLE) for GLLAMMs. The methodology is applied to estimating the effect of dietary fibre intake on coronary heart disease. We also assess the performance of estimation of regression parameters and empirical Bayes prediction of the true covariate. Normal as well as skewed covariate distributions are simulated and inference is performed based on both maximum likelihood assuming normality and NPMLE. Both estimators are unbiased and have similar root mean square errors when the true covariate is normal. With a skewed covariate, the conventional estimator is biased but has a smaller mean square error than the NPMLE. NPMLE produces substantially improved empirical Bayes predictions of the true covariate when its distribution is skewed.

Effects of prenatal alcohol exposure on the development of white matter volume and change in executive function

Prenatal alcohol exposure can cause a wide range of deficits in executive function that persist throughout life, but little is known about how changes in brain structure relate to cognition in affected individuals. In the current study, we predicted that the rate of white matter volumetric development would be atypical in children with fetal alcohol spectrum disorders (FASD) when compared to typically developing children, and that the rate of change in cognitive function would relate to differential white matter development between groups. Data were available for 103 subjects [49 with FASD, 54 controls, age range 6–17, mean age = 11.83] with 153 total observations. Groups were age-matched. Participants underwent structural magnetic resonance imaging (MRI) and an executive function (EF) battery. Using white matter volumes measured bilaterally for frontal and parietal regions and the corpus callosum, change was predicted by modeling the effects of age, intracranial volume, sex, and interactions with exposure status and EF measures. While both groups showed regional increases in white matter volumes and improvement in cognitive performance over time, there were significant effects of exposure status on age-related relationships between white matter increases and EF measures. Specifically, individuals with FASD consistently showed a positive relationship between improved cognitive function and increased white matter volume over time, while no such relationships were seen in controls. These novel results relating improved cognitive function with increased white matter volume in FASD suggest that better cognitive outcomes could be possible for FASD subjects through interventions that enhance white matter plasticity.

•

Children with prenatal alcohol exposure compared with controls longitudinally

•

Similar increases of white matter and cognition seen in both groups

•

But, brain–behavior relationships differed between groups.

•

Different relationships seen with neurobehavior and structure in exposed children

•

Results suggest interventions for white matter plasticity would be useful for FASD.

Examining the Racial Crossover in Mortality between African American and White Older Adults: A Multilevel Survival Analysis of Race, Individual Socioeconomic Status, and Neighborhood Socioeconomic Context

We examine whether individual and neighborhood socioeconomic context contributes to black/white disparities in mortality among USA older adults. Using national longitudinal data from the Americans' Changing Lives study, along with census tract information for each respondent, we conduct multilevel survival analyses. Results show that black older adults are disadvantaged in mortality in younger old age, but older black adults have lower mortality risk than whites after about age 80. Both individual SES and neighborhood socioeconomic disadvantage contribute to the mortality risk of older adults but do not completely explain race differences in mortality. The racial mortality crossover persists even after controlling for multilevel SES, suggesting that black older adults experience selective survival at very old ages. Addressing the individual and neighborhood socioeconomic disadvantage of blacks is necessary to reduce mortality disparities that culminate in older adulthood.

Review of analytical methods for prospective cohort studies using time to event data: single studies and implications for meta-analysis

Prospective cohort studies are extremely important in epidemiological research as they give direct information on the sequence of events, which can be used to demonstrate causality. They also have the advantage that many diseases can be studied simultaneously. However, they are usually very time consuming and expensive to run. In addition, practitioners of evidence-based medicine prefer to make decisions based on several studies rather than a single study, hence the need for meta-analysis. The use of meta-analyses in order to synthesize the evidence from randomized controlled trials is extremely popular in medicine and is also being utilized increasingly in epidemiology. The statistical methodology for meta-analyses of epidemiological studies is a long way behind in terms of the advances made in the methodology for randomized controlled trials. Numerous methodological issues, particularly in respect to dealing with biases inherent in these types of studies, have made the results of meta-analyses of epidemiological studies that use summary data open to criticism. This review mainly concentrates on analytical methods for prospective cohort studies that have survival outcomes. In addition, the implications for meta-analysis assuming that the analyst has access to individual participant data are also discussed. The approaches are described with respect to underlying theory and assumptions. It is hoped that this review will promote the use of these approaches in meta-analyses conducted in epidemiology as well as providing some directions for future research.