104年度導入健康風險評估科技 精進我國食品安全雄才大略計畫 食 …nehrc.nhri.org.tw/foodsafety/ref/(9)20151007.pdf ·...
Post on 10-Jul-2020
13 Views
Preview:
Transcript
104年度導入健康風險評估科技精進我國食品安全雄才大略計畫食品風險評估人才訓練課程
October 7, 2015
劑量反應評估-
基準劑量軟體教學與演練
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Dose-response assessment:Benchmark dose modeling
and practicesShu-Han You, Ph.D.
Postdoctoral Fellow National Institute of Environmental Health Sciences
National Health Research Institutes, Taiwan
2
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Learning Objects of the Course
Provide participants with training on
1. General Benchmark Dose (BMD) methods and its application to dose-response assessment
2. U.S. EPA BMD guidance
3. Use of U.S. EPA’s BMD software
This course is not intended to be a primer on basic concepts of toxicology,
nor a detailed examination of the statistical underpinnings of dose-response models.
3
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Outline
4
1. Review of Key Terminology
2. Dose-Response Assessment in Risk Analysis
3. How to Implement and Creating a Dataset in BMD analysis?
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Outline
5
1. Review of Key Terminology
2. Dose-Response Assessment in Risk Analysis
3. How to Implement and Creating a Dataset in BMD analysis?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 6
Review of Key Terminology 劑量反應關係
Dose-Response RelationshipThe relationship between a quantifiedexposure (dose) and the proportion ofsubjects in incidence and /or in degree ofchange (response).
基準反應Benchmark Response (BMR)
A predetermined change in the responseof an adverse effect relative to thebackground response of this effect.
沒有可觀察到的不良影響值
No Observed Adverse Effect Level (NOAEL)
The highest exposure level at which there areno biologically significant increases in theseverity of adverse effects between the exposedpopulation and its appropriate control group.
基準劑量Benchmark Dose (BMD)
A dose of a substance that when ingestedproduces a predetermined change in theresponse of an adverse effect relative tothe background response of this effect.
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Outline
7
1. Review of Key Terminology
2. Dose-Response Assessment in Risk Analysis
3. How to Implement and Creating a Dataset in BMD analysis?
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Dose-Response Assessment in Risk Analysis
8
Research
Risk management
Risk assessment
1. Hazard identification
2. Dose-response assessment3. Exposure assessment4. Risk characterization
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Dose-Response Curve
9
To establish a cause-effect relationship between exposure to a toxic substance (dose administered) and an observed effect (response)
in order to determine a safe exposure level
Dose
Re
spo
nse
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Threshold dose (ThD)閾值劑量
10
Dose
Re
spo
nse
閾值劑量為低於檢測無不利效應觀察值
(ThD below which no adverse effects are detected)
NOAEL
ThD approximated by a NOAEL (No Observed Adverse Effect Level,沒有可觀察到的不良影響值)
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Non-Threshold 非閾值劑量
11
Dose
Re
spo
nse
Non-Thresholdex: genotoxic carcinogens …
Threshold
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Benchmark Dose (BMD) Approach (1)
NOAEL approach used only single points, shape of dose-response is ignored.
BMD is calculated from the curve fitted to the dose-response data, so all information is used.
BMD is not replacement for NOAEL.
12
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Benchmark Dose (BMD) Approach (2)1. A mathematical
model is applied to the experimental data to produce a dose-response curve.
2. BMR is defined as 10% (or 5%, or 1%).
13Dose
% R
esp
on
se
BMRBMD
Animal model: Mouse, rat, etc
Mathematical model
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Outline
14
1. Review of Key Terminology
2. Dose-Response Assessment in Risk Analysis
3. How to Implement and Creating a Dataset in BMD analysis?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 15
Scenario 1 (Problem formulation)(Chen H, et al. 2015. Environmental Toxicology and Pharmacology)
Polychlorinated biphenyls(PCB)Developing country: China
Recycled electronic waste ↑
Source from http://www.ecy.wa.gov/programs/swfa/pbt/pcb.html
If the current toxicological threshold will really shield people
from harm of PCBs toxicity (Thyroid disruption)?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 16
Groups Number Measurement:Thyroid follicular epithelium/colloid ratio (%)
0 10 0.67
0.1 10 0.64
1 10 0.99
5 10 1.2
10 10 1.72
Table 1. Data in PCBs-treated ovariectomized SD rats for using BMD approach. (Chen et al., 2015)
1) Manually enter these data (Table 1)2) Run the Logistic, Loglogistic, and Weibull (Dichotomous) models 3) Output your results and check the answer Threshold
食品風險評估人才訓練課程:食品安全風險評估之實務演練 17
Creating a dataset
Running an specific model
Opening output and
plot files
How to implement and creating a dataset in BMDS?
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Running an specific model
Opening output and
plot files after analysis
Creating a dataset
18
How to implement and creating a dataset in BMDS?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 19
How to implement and creating a dataset in BMDS?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 20
Open new dataset
CLICK
1
2
3
食品風險評估人才訓練課程:食品安全風險評估之實務演練 21
Enter Data Manually
Enter Data Manually
Groups Number Measurement: Thyroid follicular epithelium/colloid
ratio (%)
0 10 0.67
0.1 10 0.64
1 10 0.99
5 10 1.2
10 10 1.72
00.1
15
10
10101010
10
0.670.640.99
1.2
1.72
食品風險評估人才訓練課程:食品安全風險評估之實務演練 22
Renaming column headers
CLICK RIGHT MOUSE BUTTON→RENAME (ex: dose, N, etc.)
0.115
10
1010
10
0.670.640.99
1.2
1.72Groups Number Measurement:
Thyroid follicular epithelium/colloid ratio (%)
0 10 0.67
0.1 10 0.64
1 10 0.99
5 10 1.2
10 10 1.72
食品風險評估人才訓練課程:食品安全風險評估之實務演練 23
How to implement and creating a dataset in BMDS?
Opening output and
plot files after analysis
Creating a dataset
Running an specific model
食品風險評估人才訓練課程:食品安全風險評估之實務演練 24
Running an specific model
Select a Model Type
食品風險評估人才訓練課程:食品安全風險評估之實務演練 25
Running an specific model
Select a Model Name
食品風險評估人才訓練課程:食品安全風險評估之實務演練 26
Biological interpretation
Policy decision
Otherwise
• Saturabel processes demonstrating Miichaelis-Mentenkinetics (Hill model)
• Two-stage clonal expansion model
• U.S. EPA’s IRIS program uses the multistage model for cancer data
• Whether various models mathematically describe the data
Selection of an Appropriate Dichotomous Model
食品風險評估人才訓練課程:食品安全風險評估之實務演練 27
Running an specific model
Proceed to option screen
食品風險評估人才訓練課程:食品安全風險評估之實務演練 28
Option Screen - Step 1: Selecting Colum assignments
Groups Number Thyroid follicular epithelium/colloid ratio (%)
0 10 0.67
0.1 10 0.64
1 10 0.99
5 10 1.2
10 10 1.72
食品風險評估人才訓練課程:食品安全風險評估之實務演練
To prevent biologically implausible results(U.S. EPA Recommendations)
Background do not set to zero unless biologically justifiable
Multistage β restrict to be positive
Power and slope restrict to be 1 or greater
29
Option Screen - Step 2: Selecting parameter assignments
Logistic model Weibull model Multistage model
食品風險評估人才訓練課程:食品安全風險評估之實務演練 30
Option Screen - Step 3: Selecting Other assignments
Default
An extra risk of 10% is recommended as a standard by the IRIS (not default)
食品風險評估人才訓練課程:食品安全風險評估之實務演練 31
How to implement and creating a dataset in BMDS?
Running an specific model
Creating a dataset
Opening output and
plot files after analysis
食品風險評估人才訓練課程:食品安全風險評估之實務演練 32
Opening output and plot files
2. Fit statistics
3. BMD & BMDL Estimates
1. Parameter estimates
食品風險評估人才訓練課程:食品安全風險評估之實務演練 33
Opening output and plot files
2. Fit statistics
3. BMD & BMDL Estimates
1. Parameter estimates
Does the Model Fit the Data?
Global Goodness-of-fit p-value (p > 0.1)(p數值越小表示擬合越不好)
(Global fit)
食品風險評估人才訓練課程:食品安全風險評估之實務演練 36
Opening output and plot files
2. Fit statistics
3. BMD & BMDL Estimates
1. Parameter estimates
(Local fit)
Scaled residuals (Absolute value < 2.0)
Does the Model Fit the Data?
食品風險評估人才訓練課程:食品安全風險評估之實務演練 38
Opening output and plot files
2. Fit statistics
3. BMD & BMDL Estimates
1. Parameter estimates
食品風險評估人才訓練課程:食品安全風險評估之實務演練 39
Groups Number Thyroid follicular epithelium/colloid ratio (%)
0 10 0.67
0.1 10 0.64
1 10 0.99
5 10 1.2
10 10 1.72
Table 1. Data in PCBs-treated ovariectomized SD rats for using BMD approach. (Chen et al., 2015)
ModelGoodness-of-fit
BMD BMDLp-value AIC
Logistic
Loglogistic
Weibull
Run BMD and Output your results
食品風險評估人才訓練課程:食品安全風險評估之實務演練 40
Model Goodness-of-fit BMD BMDL
p-value AIC
Logistic 0.9952 36.7 9.35337 4.3235
Loglogistic 0.9969 36.7 9.06215 2.23557
Weibull 0.9968 36.7 9.08583 2.54285
1. More than one model p>0.1 & lowest AIC will result in an acceptable fit to the data.
2. Consider using the lowest BMDL, if BMDL estimates from acceptable models are not sufficiently close, indicating model dependence.
• 通常BMD數值應不能大於BMDL 3倍
>3-fold
>3-fold
Check the Answer
/
/
食品風險評估人才訓練課程:食品安全風險評估之實務演練
Thank you
for your listening
42
top related