Application Of Latent Variable Modeling And Related Techniques To The Analysis Of Toxicological Data
Okunola, Idunnuoluwa 1992-
Background: Soil contamination poses a significant problem in Canada because of either current or potential adverse impact on human health and the environment. Petroleum hydrocarbons (PHCs) are the most common sources of soil contamination in Canada and therefore the provision of remediation targets for such contaminated soils are of great concern to toxicologists. Objective: This research project provides toxicologists with an alternative method for provisional remediation targets based on readily measured environmental variables without requirements for extensive toxicological testing. This study allows us to determine if models describing the relationships among soil characteristics, contaminant concentrations, and species responses could be used to predict these effects in soil when the contaminant concentrations and soil characteristics were known. Methods: In this study, we used statistical methods to describe the relationship among soil characteristics, contaminant concentrations, and species responses, and how these can be used to predict toxicity in soils contaminated with petroleum hydrocarbons. Structural Equation Modeling (SEM) is a useful analysis tool that can be used to analyze these covariates, while accounting for those covariates that are intercorrelated, which are usually problems in current methods. Confirmatory factor analysis (CFA) under SEM was carried out using the lavaan package in R to estimate the measurement model which specifies the relationship between covariates and their latent factors and any inter-correlations between the covariates. A structural model was also analyzed to estimate the relationships between the latent factors. Non-linear procedures were carried out to quantify the relationships between PHC contaminant concentrations and the observed species responses to provide an estimate of the concentration at which there is a particular percentage change (ICp; where p stands for the percentage change) in biological function for each endpoint (growth, reproduction, mass, shoot length, root length). Lognormal, exponential and gamma distributions were fit to the estimated IC25 and IC50 values using the ”fitdist” function in the “fitdistplus” package of R software. A lognormal distribution gave the best fit to the IC25 and IC50 values. Results: The CFA was carried out on different models specified based on theoretical knowledge and the model with the best fit was identified. This CFA model specified that the masses (or other similar responses like size) of one species are indicators of the response of that species to the PHC contaminant in the soil and this response contributes to the aggregate response of all the other species to the PHC in the soil. Soil properties were added to this model to identify how some common soil properties affect toxicity. The amount of clay and the pH of the soil were found to be significant predictors of the aggregate response of the species. PHC concentrations were also found to be a significant cause of the aggregate response. IC25 and IC50 values were estimated for the two different study sites included in the dataset. The remediation guidelines for the PHC contaminated soils according to the IC25 values were estimated as 452.76 ± 50.38 mg/kg for site 1 and 234.93 ± 394.78 mg/kg for site 2. Therefore, PHC concentrations above these levels will be of great concern. Conclusion: According to (CCME, 1996), the development of site-specific remediation objectives for PHC contaminated sites is a critical stage, and using current methods, requires extensive site-specific testing. This study demonstrated the utility of SEM in describing toxicity effects and most importantly the use of CFA in aggregating species endpoints to describe their joint response to PHC contamination. This method provides an alternative to current methods to estimate IC25 and IC50 values directly from the estimated aggregate species response. These values will then serve as remediation targets for toxicologists to use in risk assessments.
DegreeMaster of Science (M.Sc.)
DepartmentSchool of Public Health
SupervisorKhan, Shahedul; Lamb, Eric G.
CommitteeStephenson, Gladys; Roy, Chanchal; Szafron, Michael; Lim, June
Copyright DateOctober 2016