Estimation in generalized linear models under censored covariates with an application to MIREC data
- DOI
- Language of the publication
- English
- Date
- 2018-08-30
- Type
- Article
- Author(s)
- Lee, Wan-Chen
- Sinha, Sanjoy K.
- Arbuckle, Tye E.
- Fisher, Mandy
- Publisher
- Wiley
Abstract
In many biological experiments, certain values of a biomarker are often nondetectable due to low concentrations of an analyte or the limitations of a chemical analysis device, resulting in left-censored values. There is an increasing demand for the analysis of data subject to detection limits in clinical and environmental studies. In this paper, we develop a novel statistical method for the maximum likelihood estimation in generalized linear models with covariates subject to detection limits. Simulations are carried out to study the relative performance of the proposed estimators, as compared to other existing estimators. The proposed method is also applied to a real dataset from the Maternal-Infant Research on Environmental Chemicals cohort study, where we investigate how different chemical mixtures affect the health outcomes of infants and pregnant women.
Plain language summary
Health Canada is responsible for the assessment and management of health risks to Canadians associated with exposure to products and chemicals in the environment. As part of its activities, Health Canada measured levels of certain environmental chemicals in a large group of Canadian women during pregnancy. As often happens in such studies, some samples had undetectable levels of some chemicals -- a situation where the level of the chemical, by convention, is reported as “below the limit of detection” (denoted “below LOD” or “LOD”). The simplest method of dealing with LOD observations is to remove all LOD observations and continue statistical analysis based on those samples with measureable levels. Such analysis is not generally recommended because useful information showing low exposure levels is lost. Another way to handle the issue is to replace LOD observations by a constant, like LOD or LOD/2, and then apply statistical analysis. Unfortunately, the results from the data with substituted exposure levels are often biased. Here, Health Canada proposed a new method to deal with the issue of LOD observations. A simulation study was conducted to show that the results using this method were not biased and were close to the expected value. The method was then applied to the Maternal-Infant Research on Environmental Chemicals (MIREC) Study Cohort to study the potential for synergistic effects of simultaneous exposures to two environmental chemicals on health outcomes in pregnant women. Preliminary results suggest simultaneous exposure to certain pairs of environmental chemicals affected pregnant women and their infants in ways that could not be explained by simply adding the individual effects of each chemical. Future work will expand this method to examine potential synergistic effects of simultaneous exposure to several chemicals thus allowing Health Canada to better estimate the health effects of exposures to multiple environmental chemicals.
Subject
- Health,
- Health and safety