Zeitschrift für Biometrie und Biostatistik

In health related research it is common to have multiple outcomes of interest in a single study. These outcomes are often analysed separately, ignoring the correlation between them. One would expect that a multivariate approach would be a more efficient alternative to individual analyses of each outcome. Surprisingly, this is not always the case. In this article we discuss different settings of linear models and compare the multivariate and univariate approaches. We show that for linear regression models, the estimates of the regression parameters associated with covariates that are shared across the outcomes are the same for the multivariate and univariate models while for outcome-specific covariates the multivariate model performs better in terms of efficiency.

Background: Prenatal screening combines biochemical and biometric markers into a risk estimate for a particular adverse outcome, e.g. the birth of a child with Down syndrome. The statistical calculations are complicated. We describe a simple graphical method to perform risk estimation in the case of two biochemical markers, to assess the consequences of changes in gestational dating of the pregnancy and to perform quality control. Materials and methods: We used the formulae for the Normal distribution to establish the expression for fractal ellipses, i.e. contour ellipses describing a certain fractile of the total distribution. This expression was used to establish mathematical expressions for curves describing two-dimensional pairs of analytes giving the same likelihood-ratio, i.e. iso likelihood-ratio curves. Results: The fractal ellipses provide an overview of marker distributions that allow for an easy control of empirical marker distributions. The iso likelihood-ratio curves provide a relation between likelihood-ratio and marker values. They can be used for assessment of the consequences of changes in gestational age and introduction of truncation limits on markers. Conclusions: Fractal ellipses and iso likelihood-ratio curves can be used to make software-independent calculations and modifications of risk in prenatal screening and quality control of an ongoing screening program.

Primary open angle glaucoma (POAG) is a chronic, progressive, irreversible, and potentially blinding optic neuropathy. The risk of blindness due to progressive visual field (VF) loss varies substantially from patient to patient. Early identification of those patients destined to rapid progressive visual loss is crucial to prevent further damage. In this article, a latent class growth model (LCGM) was developed to predict the binary outcome of VF progression using longitudinal mean deviation (MD) and pattern standard deviation (PSD). Specifically, the trajectories of MD and PSD were summarized by a functional principal component (FPC) analysis, and the estimated FPC scores were used to identify subgroups (latent classes) of individuals with distinct patterns of MD and PSD trajectories. Probability of VF progression for an individual was then estimated as weighted average across latent classes, weighted by posterior probability of class membership given baseline covariates and longitudinal MD/PSD series. The model was applied to the participants with newly diagnosed POAG from the Ocular Hypertension Treatment Study (OHTS), and the OHTS data was best fit by a model with 4 latent classes. Using the resultant optimal LCGM, the OHTS participants with and without VF progression could be accurately differentiated by incorporating longitudinal MD and PSD.

Background: For cohort and cross-sectional studies, the risk ratio (RR) is the preferred measure of effect rather than an odds ratio (OR), especially when the outcome is common (>10%). The log-binomial (LB) and Poisson models are commonly used to estimate the RR; the OR estimated using logistic regression is often used to approximate the RR when the outcome is rare. However, regardless of the prevalence of the outcome, logistic regression predicted exposed and unexposed risks may be used to estimate the RR. Because maximum likelihood estimation is used to fit the logistic model, estimation of the SE of the RR is difficult. Methods: To overcome difficulty in estimation of the SE of the RR and provide a flexible framework for modeling, we developed a Bayesian logistic regression (BLR) model to estimate the RR, with associated credible interval (CIB). We applied the BLR model to a large hypothetical cross-sectional study with categorical variables and to a small hypothetical clinical trial with a continuous variable for which the LB method did not converge. Results of the BLR model were compared to those from several commonly used RR modeling methods. Results: Our examples illustrate the Bayesian logistic regression model estimates adjusted RRs and 95% CIBs comparable to results from other methods. Adjusted risks and risk differences were easily obtained from the posterior distribution. Conclusions: The Bayesian logistic regression modeling approach compares favorably with existing RR modeling methods and provides a flexible framework for investigating confounding and effect modification on the risk scale.

A common practice in gene expression studies is to use ‘housekeepers’, i.e., genes expected to be expressed at relatively constant levels across experimental conditions, to normalize data. The process is to divide an expression value by some composite of one or more stable housekeepers to remove the effect of processing and nuance variables. Despite its reverence and widespread use, we argue that this approach is fundamentally flawed on multiple levels. The outcome of housekeeper normalization is a set of ratio variables which are not amenable to many standard statistical tests. There are no universal housekeeper genes and even within specific cohorts proposed housekeeper genes often fail to replicate. Furthermore, there is also no single agreed upon algorithm for performing housekeeper normalization or agreement regarding what constitutes a good housekeeper. We urge researchers to consider the use of alternative methodologies in their research.

The main objectives of this study is modelling acute respiratory infection symptoms among under five children and to investigate how different explanatory variables measured at different level of hierarchical structures affects symptoms of ARI. This study used Ethiopian Demographic and Health Survey (EDHS) 2011 data, collected for 9625 children under five years old in Ethiopia and children are nested within eleven geographical regions. Binary logistic regression analysis and multilevel models were employed to predict the outcome. The study revealed that mothers educational level, age of children, number of children, mothers occupational status, supplementation of vitamin A, source of drinking water, type of toilet facility and wealth index of family were found to be the most important factors. And, the final model, random coefficient multilevel logistic regression suggests that there exists considerable differences in the ARI symptoms among under five children across the regions. It indicates that the variance of random component related to the random term were found to be statistically significant, implying that their is differences in the ARI symptoms for children across regions. The study suggests that improve mothers educational level in all of areas in order to address the problem through improving their income earning capacity, improve access of safe drinking water and the researcher who want to conduct ARI symptoms among children under five using EDHS data set should use multilevel model than classical regression models.

Between January and June 2012, a moderate-sized Legionella pneumophila serogroup 1 (Lp1) outbreak occurred in Auckland, New Zealand, which involved 19 cases, with two deaths. Initial investigation did not reveal a common source. Poorly maintained cooling towers were a likely cause, and mass shock dosing of all such towers with biocide was undertaken in April 2012 and repeated after an almost identical outbreak in the autumn of 2013. Our aim was to assess whether shock dosing of towers affected disease incidence. A time-series analysis, using regression discontinuity, of the notified Lp1 cases from 2007 to October 2014 was carried out. A total of 84 out of 92 cases of Lp1 were available for analysis. Seasonal trend decomposition showed an excess of cases in the autumn of 2012 and 2013, with a decline in 2014. Poisson regression showed an average log-linear annual increase in monthly notifications by 40% (95% confidence interval (CI): 17% to 68%), with an average 46% decline (95% CI: 74% decrease to 13% increase) comparing cases before April 2012 to those that occurred afterward. In dispersed outbreaks in urban settings, we conclude that this study supports mass shock dosing of cooling towers to limit disease occurrence.

Exact locations of patients, primary residences at the time of diagnoses are routinely collected as part of the TB surveillance program to ability clusters and detect disease outbreaks Tuberculosis early is important in order to decrease morbidity and mortality through timely implementation of disease prevention and control measures. It has been shown for syndromic surveillance data that when exact geographic coordinates of individual patients are used, higher detection rates and accuracy are achieved compared to when data are aggregated into administrative regions such as zip codes and census tracts. Many national, state, and local health departments are launching disease surveillance systems with daily analyses of hospital emergency department visits, ambulance dispatch calls, or pharmacy sales for which population-at-risk information is inaccessible or inappropriate.

Asthma and obesity are two significant public health problems that both originate in early childhood and have shared risk factors and manifestations. Studies suggest a strong association between asthma development and subsequent accelerated weight gain. Children are diagnosed with asthma in early childhood and are often exposed to factors associated with rapid weight gain. This article intends to demonstrate an innovative application of the piecewise mixed effects model to characterize the difference in the temporal rate of change in BMIz, the standardized scores of body mass index and weight-for-length that measure weight status, before and after asthma diagnosis in children younger than 5 years. The data consist of unique sequences from 1194 children's clinic visits during the first 5 years of life. We used a knot at the time of diagnosis and detected a differential weight-gain pattern before and after asthma diagnosis. The pre- and post-asthma-diagnosis weight-gain patterns further differ by sex and race-ethnicity. After asthma diagnosis, female children showed a higher increase in the rate of change in BMIz than males. Non-Hispanic African Americans and Hispanics had higher post-diagnosis rates of change in BMIz than Caucasians. The differential weight-gain patterns between male and female children were mainly contributed by Caucasian children. These findings could have important implications in the clinical care of children after asthma diagnosis.

Purpose: The purpose of this paper is to examine the sensitivity and mean sojourn time specific lead time distribution in cancer screening trials when lifetime is a random variable in order to explore possible optimal initial age at screening and screening frequency. Methods: Summarized methods from Wu et al. (2012). Simulation was used in order to estimate the distribution of the lead time for a hypothetical individual with a future screening schedule. The lifetime distribution used comes from the Social Security Administration’s actuarial life tables. The lead time distribution was then calculated based on a loglogistic sojourn time distribution with two mean sojourn times (2, and 10 years), using three different initial screening ages, t0=40, 50, 60, different screening sensitivities (0.3 and 0.5 for men; 0.8 and 0.9 for women), and two different screening frequencies, one and two years for both men and women. Results: Smaller time intervals between screenings yield a smaller probability of no early detection and a greater expected lead time.