World Congress on Risk 2015
19-23 July, 2015, Singapore

Online Program

Session Schedule & Abstracts

* Disclaimer: All presentations represent the views of the authors, and not the organizations that support their research. Please apply the standard disclaimer that any opinions, findings, and conclusions or recommendations in abstracts, posters, and presentations at the meeting are those of the authors and do not necessarily reflect the views of any other organization or agency. Meeting attendees and authors should be aware that this disclaimer is intended to apply to all abstracts contained in this document. Authors who wish to emphasize this disclaimer should do so in their presentation or poster. In an effort to make the abstracts as concise as possible and easy for meeting participants to read, the abstracts have been formatted such that they exclude references to papers, affiliations, and/or funding sources. Authors who wish to provide attendees with this information should do so in their presentation or poster.

Common abbreviations

Tuesday 21-07-2015

Food Safety Assessment

Room: Breakthrough   11:00–12:30

Chair(s): Kuen-Yuh Wu

2    Establishment of food intake databank for Taiwan. Chang H.Y.    (210)

Abstract: Background: Recently, many countries are establishing their food data banks. The purpose is to provide timely data for health risk assessment. Food items also change over time, many new foods and additions emerge. The old data might not fit current situation. The need of exposure assessment is to know the amount of food intake, not nutrient. Method: This study renewed food classification and update intakes data constantly. It is hoped to establish a databank for food intakes of the nation. Nutrition and Health Survey in Taiwan (NAHSIT) focus on nutrient intakes. However, it is the only database with food items and amount of intakes based on nutrients. It does not meet the need of exposure assessment. Results: We reclassified food to accommodate the needs of regulation and risk assessment. The final classification included three levels. We also estimated the population mean and SD of intakes after accounting for sampling scheme. We calculated the results with mean, SD, minimum, and maximum for each food items as well as food groups. Results included (1) re-classified food into three into 3 level food groups ; (2) re-estimating the age*sex specific intakes of each categories; (3) estimating mean and SD per capita for each food and categories of food raw weights and cooked weights. The estimations are available in the website Keyword:Food intakes, Nutritional and Health Survey in Taiwan (NAHSIT)

3    Benchmark Dose Calculation for Ordered Categorical Responses with Multiple Endpoints. Chen C.C.    (209)

Abstract: The benchmark dose (BMD) approach for exposure limit in the risk assessment of cancer and non-cancer endpoints is well-established, which is often based on dose-response modeling of the most critical or the most sensitive outcome. However, the most critical endpoint may not be available and the most sensitive endpoint may not necessarily be representative of the overall toxic effects. In this paper, to calculate BMD in the case of multivariate responses, we categorize the measurements of each endpoint into ordinal severity effects such as none, mild, adverse, and severe, so that multiple responses can be accommodated and mutual comparisons can be made simultaneously. A latent ordered categorical variable is obtained by expressing it as a common factor loading for the multinomial distribution parameters of each of the multivariate ordered categorical responses, similar to the principal components analysis (PCA) approach of Budtz-Jørgensen (2007) for continuous variables. We then apply the BMD calculation method for ordered categorical response of Chen and Chen (2014) to the resultant latent variable for the corresponding benchmark dose. A Bayesian statistical approach is employed using Markov chain Monte Carlo simulation to obtain the latent ordinal variable and the model parameter estimates. A carcinogenesis study of acrylamide in rats with multivariate endpoints of different severity levels is analyzed for illustration. KEY WORDS: Dirichlet distribution; factor loading; latent variable; Markov chain Monte Carlo simulation; multinomial distribution; principal component analysis

4    Probabilistic Assessment of Health Risk on Ochratoxin A with Bayesian Statistics Markov Chain Monte Carlo Simulation. Chiang S.Y.    (214)

Abstract: Ochratoxin A (OTA) is a mycotoxin produced by several fungal species of the genera Penicillium and Aspergillus, which often contaminate food commodities, including cereals, coffee, beer, wine, dried fruits and cacao. OTA caused nephrotoxicity, carcinogenicity, teratogenicity and immunotoxicity to rodents and might be associated with Balkan endemic nephropathy (BEN) and chronic interstitial nephropathy (CIN). OTA also has genotoxicity in mammalian cells and mice. Daily exposures to OTA through food consumption have been of great concerns. Therefore, the objective of this study was to conduct a probabilistic risk assessment on OTA with the Bayesian statistics Markov chain Monte Carlo (BS-MCMC) simulation to overcome insufficient data. The OTA data were cited from reports released by Taiwan Food and Drug Administration (TFDA), and the intake rates of foods were cited from the Taiwan National Food Consumption Database. These data were used as prior information. The posterior distributions of OTA residues, daily intake, cancer risk, and margin of exposures (MoE) for adults in Taiwan were assessed with the BS-MCMC simulation by using the OpenBUGS software. The cancer slope factor of OTA was assessed with the Benchmark dose software and extrapolated to human according to the bodyweight ratio, and MoE was equal to BMDL10 divided by daily OTA exposures. The mean cancer risk for the male adults in Taiwan is 7.55×10-06 and the 95% upper limit is 3.09×10-05 which is greater than 6.68×10-06 and 3.92×10-05 for the female adults. The mean MoE and the 95% lower limit for the male adults is 13,229 and 5154 which is less than 14,966 and 6221 for the female adults. These results demonstrate limitation of insufficient data can be improved by using the BC-MCMC method to reduce uncertainties in traditional Monte Carlo simulation. The distributions of cancer risk and MoE suggest that TFDA should pay attention on the management and the storage of these foods. Key words: Ochratoxin A; probabilistic risk assessment; Bayesian statistics Markov chain Monte Carlo simulation

5    Probabilistic Risk Assessment by Using Bayesian Statistics-Markov Chain Monte Carlo Simulation. Wu K.Y.    (215)

Abstract: US EPA already adopted probabilistic risk assessment (PRA) for decision making. Previously, PRA was conducted by mainly using the Monte Carlo (MC) simulation, which frequently requires either empirical or probability distributions of parameters to simulate the distribution of lifetime daily dose. The simulation results will be valid if only the input parameters, data and assumptions are valid. In practice, risk assessors frequently suffered from insufficient data to fit distributions for some parameters, especially concentrations and intake rates, or even worse spotted data hinder completion of an assessment, such as a large proportion of residue data below detection limit. In order to reduce uncertainty due to insufficient data, the Bayesian statistics Markov chain Monte Carlo (MCMC) simulation was applied to perform PRA. The limited data available were used as prior information. Markov chain Monte Carlo simulation was performed with the WinBUG to achieve the posterior distributions of parameters and health risk. Several examples will be presented in this meeting; such as assessment of lifetime cancer risk for ochratoxin A (OTA) in foods, assessment of lifetime cancer risk for aflatoxin B1 in food (only few data greater than regulations were available), assessment of lifetime cancer risk for acrylamide in high-temperature processed foods with high uncertainty in residue and intake rate data, and even the assessment of DDT and DDE cancer risk by fitting a multimedia model. With limited data available, the posterior distributions of parameters and health risk theoretically converge to corresponding representative distributions for the study population so that quality of risk assessment may be improved without additional investment of resources to collect data.

[back to schedule]