Semiparametric regression models for a right-skewed outcome subject to pooling

Local polynomial regression for pooled response data

We propose local polynomial estimators for the conditional mean of a continuous response when only pooled response data are collected under different pooling designs. Asymptotic properties of these estimators are investigated and compared. Extensive simulation studies are carried out to compare finite sample performance of the proposed estimators under various model settings and pooling strategies. We apply the proposed local polynomial regression methods to two real-life applications to illustrate practical implementation and performance of the estimators for the mean function.

Flexible modeling of ratio outcomes in clinical and epidemiological research

In medical studies one frequently encounters ratio outcomes. For modeling these right-skewed positive variables, two approaches are in common use. The first one assumes that the outcome follows a normal distribution after transformation (e.g. a log-normal distribution), and the second one assumes gamma distributed outcome values. Classical regression approaches relate the mean ratio to a set of explanatory variables and treat the other parameters of the underlying distribution as nuisance parameters. Here, more flexible extensions for modeling ratio outcomes are proposed that allow to relate all the distribution parameters to explanatory variables. The models are embedded into the framework of generalized additive models for location, scale and shape (GAMLSS), and can be fitted using a component-wise gradient boosting algorithm. The added value of the new modeling approach is demonstrated by the analysis of the LDL/HDL cholesterol ratio, which is a strong predictor of cardiovascular events, using data from the German Chronic Kidney Disease Study. Particularly, our results confirm various important findings on risk factors for cardiovascular events.

Exposure to Persistent Organic Pollutants and Birth Characteristics: The Upstate KIDS Study

BACKGROUND

Prenatal exposure to persistent organic pollutants (POPs) may be associated with obesogenic effects in offspring. Our study is the first to investigate associations between concentrations of POPs from newborn dried blood spots (DBS) and birth characteristics.

METHODS

Concentrations of 10 polychlorinated biphenyl congeners (PCBs), polybrominated diphenyl ether-47 (PBDE-47), and p,p'-dichlorodiphenyldichloroethylene (p,p'-DDE) were measured from DBSs collected at birth from 2,065 singleton infants. DBS samples were pooled in groups of five and assayed together to reach limits of detection. Differences in risk of large for gestational age (LGA, defined as >90th percentile of birth weight for sex and gestational age), small for gestational age (SGA, <10th), and preterm birth (gestational age <37 weeks) were estimated using logistic regression per unit (ng/ml) increase in concentration of each chemical, adjusting for individual-level covariates, including maternal age, race/ethnicity, prepregnancy BMI, education, parity, smoking, and infant sex while assuming a gamma distribution and using multiple imputation to account for pools.

RESULTS

There were 215 (11.3%) singletons born LGA, 158 (7.5%) born SGA, and 157 (7.6%) born preterm. Higher concentrations of POPs were positively associated with slightly higher risk of LGA and higher birth weight.

CONCLUSIONS

Relationships between POPs measured in newborn DBS and birth size were mixed. Pooled analysis methods using DBS could address challenges in limits of detection and costs for population-based research.

Determination of Varying Group Sizes for Pooling Procedure

Pooling is an attractive strategy in screening infected specimens, especially for rare diseases. An essential step of performing the pooled test is to determine the group size. Sometimes, equal group size is not appropriate due to population heterogeneity. In this case, varying group sizes are preferred and could be determined while individual information is available. In this study, we propose a sequential procedure to determine varying group sizes through fully utilizing available information. This procedure is data driven. Simulations show that it has good performance in estimating parameters.

Estimating relative risk of a log-transformed exposure measured in pools

Pooling biospecimens prior to performing laboratory assays is a useful tool to reduce costs, achieve minimum volume requirements and mitigate assay measurement error. When estimating the risk of a continuous, pooled exposure on a binary outcome, specialized statistical techniques are required. Current methods include a regression calibration approach, where the expectation of the individual-level exposure is calculated by adjusting the observed pooled measurement with additional covariate data. While this method employs a linear regression calibration model, we propose an alternative model that can accommodate log-linear relationships between the exposure and predictive covariates. The proposed model permits direct estimation of the relative risk associated with a log-transformation of an exposure measured in pools. Published 2016. This article is a U.S. Government work and is in the public domain in the USA.