Mei Cheng Wang

Neighborhood risk factors for low birthweight in Baltimore: a multilevel analysis.

OBJECTIVES Past research on low birthweight has focused on individual-level risk factors. We sought to assess the contribution of macrolevel social factors by using census tract-level data on social stratification, community empowerment, and environmental stressors. METHODS Census tract-level information on social risk was linked to birth certificate records from Baltimore, Md, for the period 1985 through 1989. Individual level factors included maternal education, maternal age, medical assistance health insurance (Medicaid), and trimester of prenatal care initiation. Methods of multilevel modeling using two-stage regression analyses were employed. RESULTS Macrolevel factors had both direct associations and interactions with low birthweight. All individual risk factors showed interaction with macrolevel variables; that is, individual-level risk factors for low birthweight behaved differently depending upon the characteristics of the neighborhood of residence. For example, women living in high-risk neighborhoods benefited less from prenatal care than did women living in lower-risk neighborhoods. CONCLUSIONS Multilevel modeling is an important tool that allows simultaneous study of macro- and individual-level risk factors. Multilevel analyses should play a larger role in the formulation of public health policies.

Nonparametric Estimation from Cross-Sectional Survival Data

Abstract In many follow-up studies survival data are often observed according to a cross-sectional sampling scheme. Data of this type are subject to left truncation in addition to the usual right censoring. A number of characteristics and properties of the product-limit estimate, for left-truncated and right-censored data, have been explored and found to be similar to those of the Kaplan-Meier estimate. Under the stationarity assumption, however, it is believed that an alternative estimate has much better efficiency. In this article the conditional maximum likelihood estimate (MLE) property of the product-limit estimate is visited. The non-parametric MLE of the truncation distribution is derived. Use of this estimate includes testing the stationarity assumption, estimating the proportion of truncated data, and other applications in prevalent cohort studies. The analysis of the estimation is based on a “working data” approach. The asymptotic properties of the proposed estimates are developed through nonpar...

Regionally-specific diffusion tensor imaging in mild cognitive impairment and Alzheimer's disease

BACKGROUND Diffusion tensor imaging (DTI) studies have shown significant cross-sectional differences among normal controls (NC) mild cognitive impairment (MCI) and Alzheimers disease (AD) patients in several fiber tracts in the brain, but longitudinal assessment is needed. METHODS We studied 75 participants (25 NC, 25 amnestic MCI, and 25 mild AD) at baseline and 3 months later, with both imaging and clinical evaluations. Fractional anisotropy (FA) was analyzed in regions of interest (ROIs) in: (1) fornix, (2) cingulum bundle, (3) splenium, and (4) cerebral peduncles. Clinical data included assessments of clinical severity and cognitive function. Cross-sectional and longitudinal differences in FA, within each ROI, were analyzed with generalized estimating equations (GEE). RESULTS Cross-sectionally, AD patients had lower FA than NC (p<0.05) at baseline and 3 months in the fornix and anterior portion of the cingulum bundle. Compared to MCI, AD cases had lower FA (p<0.05) in these regions and the splenium at 0 and 3 months. Both the fornix and anterior cingulum correlated across all clinical cognitive scores; lower FA in these ROIs corresponded to worse performance. Over the course of 3 months, when the subjects were clinically stable, the ROIs were also largely stable. CONCLUSIONS Using DTI, findings indicate FA is decreased in specific fiber tracts among groups of subjects that vary along the spectrum from normal to AD, and that this measure is stable over short periods of time. The fornix is a predominant outflow tract of the hippocampus and may be an important indicator of AD progression.

Analyzing Recurrent Event Data With Informative Censoring

Recurrent event data are frequently encountered in longitudinal follow-up studies. In statistical literature, noninformative censoring is typically assumed when statistical methods and theory are developed for analyzing recurrent event data. In many applications, however, the observation of recurrent events could be terminated by informative dropouts or failure events, and it is unrealistic to assume that the censoring mechanism is independent of the recurrent event process. In this article we consider recurrent events of the same type and allow the censoring mechanism to be possibly informative. The occurrence of recurrent events is modeled by a subject-specific nonstationary Poisson process via a latent variable. A multiplicative intensity model is used as the underlying model for nonparametric estimation of the cumulative rate function. The multiplicative intensity model is also extended to a regression model by taking the covariate information into account. Statistical methods and theory are developed for estimation of the cumulative rate function and regression parameters. As a major feature of this article, we treat the distributions of both the censoring and latent variables as nuisance parameters. We avoid modeling and estimating the nuisance parameters by proper procedures. An analysis of the AIDS Link to Intravenous Experiences cohort data is presented to illustrate the proposed methods.

A semiparametric model for randomly truncated data

Abstract For randomly censored data, it is known that the maximum likelihood estimate (MLE) of the survival curve is not affected by parametric assumption on the censoring variable. The Kaplan-Meier (1958) estimate is the MLE for both nonparametric and semiparametric models. For randomly truncated data, the truncation product-limit estimate is the MLE for nonparametric models. This is not the case if the truncation mechanism is parameterized, however. Specifically, let X be a generic random variable and T be the truncation variable. If the distribution of T is parameterized and the distribution of X is left unspecified, it can be shown that the truncation product-limit estimate is not the MLE for this semiparametric model, even though it is for the fully nonparametric model. In this article the MLE is characterized for the semiparametric model, and the large-sample properties of the estimate are established. The results show that, unlike censoring, the parametric information from the truncation mechanism ...

Joint modeling and estimation for recurrent event processes and failure time data

Recurrent event data are commonly encountered in longitudinal follow-up studies related to biomedical science, econometrics, reliability, and demography. In many studies, recurrent events serve as important measurements for evaluating disease progression, health deterioration, or insurance risk. When analyzing recurrent event data, an independent censoring condition is typically required for the construction of statistical methods. In some situations, however, the terminating time for observing recurrent events could be correlated with the recurrent event process, thus violating the assumption of independent censoring. In this article, we consider joint modeling of a recurrent event process and a failure time in which a common subject-specific latent variable is used to model the association between the intensity of the recurrent event process and the hazard of the failure time. The proposed joint model is flexible in that no parametric assumptions on the distributions of censoring times and latent variables are made, and under the model, informative censoring is allowed for observing both the recurrent events and failure times. We propose a “borrow-strength estimation procedure” by first estimating the value of the latent variable from recurrent event data, then using the estimated value in the failure time model. Some interesting implications and trajectories of the proposed model are presented. Properties of the regression parameter estimates and the estimated baseline cumulative hazard functions are also studied.

Statistical models for prevalent cohort data

In prospective cohort studies individuals are sometimes recruited according to a certain cross-sectional sampling criterion. A prevalent cohort is defined as a group of individuals who have a certain disease at enrollment into the study. Statistical models for the analysis of prevalent cohort data are considered when the onset or diagnosis time of the disease is known. The incident proportional hazards model, where the time scale is duration with disease, is compared to the prevalent proportional hazards model, where the fundamental time scale is follow-up time. In certain cases the time of enrollment may coincide with another event (such as the initiation of treatment). This situation is also considered and its limitations highlighted. To illustrate the methodological ideas discussed in the paper, the analysis of data from an observational study of zidovudine (ZVD) in patients with the acquired immunodeficiency syndrome (AIDS) is presented.

Improving Child and Parent Mental Health in Primary Care: A Cluster-Randomized Trial of Communication Skills Training

OBJECTIVE. We examined child and parent outcomes of training providers to engage families efficiently and to reduce common symptoms of a range of mental health problems and disorders. METHODS. Training involved three 1-hour discussions structured around video examples of family/provider communication skills, each followed by practice with standardized patients and self-evaluation. Skills targeted eliciting parent and child concerns, partnering with families, and increasing expectations that treatment would be helpful. We tested the training with providers at 13 sites in rural New York, urban Maryland, and Washington, DC. Children (5–16 years of age) making routine visits were enrolled if they screened “possible” or “probable” for mental disorders with the Strengths and Difficulties Questionnaire or if their provider said they were likely to have an emotional or behavioral problem. Children and their parents were then monitored for 6 months, to assess changes in parent-rated symptoms and impairment and parent symptoms. RESULTS. Fifty-eight providers (31 trained and 27 control) and 418 children (248 patients of trained providers and 170 patients of control providers) participated. Among the children, 72% were in the possible or probable categories. Approximately one half (54%) were white, 30% black, 12% Latino, and 4% other ethnicities. Eighty-eight percent (367 children) completed follow-up monitoring. At 6 months, minority children cared for by trained providers had greater reduction in impairment (−0.91 points) than did those cared for by control providers but no greater reduction in symptoms. Seeing a trained provider did not have an impact on symptoms or impairment among white children. Parents of children cared for by trained providers experienced greater reduction in symptoms (−1.7 points) than did those cared for by control providers. CONCLUSION. Brief provider communication training had a positive impact on parent mental health symptoms and reduced minority childrens impairment across a range of problems.

Ubiquitin Proteasome System Stress Underlies Synergistic Killing of Ovarian Cancer Cells by Bortezomib and a Novel HDAC6 Inhibitor

Purpose: Elevated metabolic activity of ovarian cancer cells causes increased ubiquitin-proteasome-system (UPS) stress, resulting in their greater sensitivity to the toxic effects of proteasomal inhibition. The proteasomes and a potentially compensatory histone deacetylase 6 (HDAC6)-dependent lysosomal pathway mediate eukaryotic protein turnover. We hypothesized that up-regulation of the HDAC6-dependent lysosomal pathway occurs in response to UPS stress and proteasomal inhibition, and thus, ovarian cancer cell death can be triggered most effectively by coinhibition of both the proteasome- and HDAC6-dependent protein degradation pathways. Experimental Design: To address this hypothesis, we examined HDAC6 expression patterns in normal and cancerous ovarian tissues and used a novel HDAC6-specific inhibitor, NK84, to address HDAC6 function in ovarian cancer. Results: Abnormally high levels of HDAC6 are expressed by ovarian cancer cells in situ and in culture relative to benign epithelium and immortalized ovarian surface epithelium, respectively. Specific HDAC6 inhibition acts in synergy with the proteasome inhibitor Bortezomib (PS-341) to cause selective apoptotic cell death of ovarian cancer cells at doses that do not cause significant toxicity when used individually. Levels of UPS stress regulate the sensitivity of ovarian cancer cells to proteasome/HDAC6 inhibition. Pharmacologic inhibition of HDAC6 also reduces ovarian cancer cell spreading and migration consistent with its known function in regulating microtubule polymerization via deacetylation of α-tubulin. Conclusion: Our results suggest the elevation of both the proteasomal and alternate HDAC6-dependent proteolytic pathways in ovarian cancer and the potential of combined inhibition of proteasome and HDAC6 as a therapy for ovarian cancer.

Analysis of Accelerated Hazards Models

Abstract The proportional hazards model for survival time data usually assumes that the covariates of interest take constant effects proportionally on an unspecified baseline hazard function. However, it may not be applicable when the assumption of constant proportionality is violated. In a two-arm randomized clinical trial, for example, the treatment is often expected to be fully effective only after a certain lag period. Some alternatives, such as the accelerated failure time model, have been developed in statistical literature. This article introduces an accelerated hazards model when there is a scale change relationship between the hazard functions. An estimating equation is proposed to estimate the parameter semiparametrically. The methodology is demonstrated within a two-sample framework. Several extensions of the model are also considered. Real clinical trial data are used to investigate the models practical use.